├── .DS_Store ├── .gitignore ├── .travis.yml ├── .vscode └── settings.json ├── CITATION.md ├── LICENSE ├── MANIFEST.in ├── Pipfile ├── Pipfile.lock ├── README.md ├── Untitled.ipynb ├── Untitled.py ├── archive └── qmnist.py ├── docs ├── DOCUMENT.md ├── document.txt └── index.html ├── image └── qwgc_logo.png ├── label.txt ├── nohup.out ├── notebook ├── Plots.ipynb ├── Plots.py ├── qwgc_tutorial.ipynb ├── qwgc_tutorial.py ├── random_walk.ipynb └── random_walk.py ├── qwgc ├── QWGC.py ├── QWGC_mix.py ├── QW_kernel.py ├── __init__.py ├── classifier │ ├── __init__.py │ ├── costfunc.py │ └── qcircuit.py ├── experiments.toml ├── preprocess │ ├── __init__.py │ ├── gparse.py │ ├── qwalk.py │ └── qwfilter.py └── utils │ └── .gitignore ├── requirements.txt ├── result.txt ├── result2.txt ├── setup.py └── test └── test_QWGC.py /.DS_Store: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qwqmlf/qwgc/d06c805023bf37a1252505f7dc6e30461d861440/.DS_Store -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | # Byte-compiled / optimized / DLL files 2 | __pycache__/ 3 | *.py[cod] 4 | *$py.class 5 | 6 | # C extensions 7 | *.so 8 | 9 | # Distribution / packaging 10 | .Python 11 | build/ 12 | develop-eggs/ 13 | dist/ 14 | downloads/ 15 | eggs/ 16 | .eggs/ 17 | lib/ 18 | lib64/ 19 | parts/ 20 | sdist/ 21 | var/ 22 | wheels/ 23 | pip-wheel-metadata/ 24 | share/python-wheels/ 25 | *.egg-info/ 26 | .installed.cfg 27 | *.egg 28 | MANIFEST 29 | 30 | # PyInstaller 31 | # Usually these files are written by a python script from a template 32 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 33 | *.manifest 34 | *.spec 35 | 36 | # Installer logs 37 | pip-log.txt 38 | pip-delete-this-directory.txt 39 | 40 | # Unit test / coverage reports 41 | htmlcov/ 42 | .tox/ 43 | .nox/ 44 | .coverage 45 | .coverage.* 46 | .cache 47 | nosetests.xml 48 | coverage.xml 49 | *.cover 50 | *.py,cover 51 | .hypothesis/ 52 | .pytest_cache/ 53 | 54 | # Translations 55 | *.mo 56 | *.pot 57 | 58 | # Django stuff: 59 | *.log 60 | local_settings.py 61 | db.sqlite3 62 | db.sqlite3-journal 63 | 64 | # Flask stuff: 65 | instance/ 66 | .webassets-cache 67 | 68 | # Scrapy stuff: 69 | .scrapy 70 | 71 | # Sphinx documentation 72 | docs/_build/ 73 | 74 | # PyBuilder 75 | target/ 76 | 77 | # Jupyter Notebook 78 | .ipynb_checkpoints 79 | 80 | # IPython 81 | profile_default/ 82 | ipython_config.py 83 | 84 | # pyenv 85 | .python-version 86 | 87 | # pipenv 88 | # According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control. 89 | # However, in case of collaboration, if having platform-specific dependencies or dependencies 90 | # having no cross-platform support, pipenv may install dependencies that don't work, or not 91 | # install all needed dependencies. 92 | #Pipfile.lock 93 | 94 | # PEP 582; used by e.g. github.com/David-OConnor/pyflow 95 | __pypackages__/ 96 | 97 | # Celery stuff 98 | celerybeat-schedule 99 | celerybeat.pid 100 | 101 | # SageMath parsed files 102 | *.sage.py 103 | 104 | # Environments 105 | .env 106 | .venv 107 | env/ 108 | venv/ 109 | ENV/ 110 | env.bak/ 111 | venv.bak/ 112 | 113 | # Spyder project settings 114 | .spyderproject 115 | .spyproject 116 | 117 | # Rope project settings 118 | .ropeproject 119 | 120 | # mkdocs documentation 121 | /site 122 | 123 | # mypy 124 | .mypy_cache/ 125 | .dmypy.json 126 | dmypy.json 127 | 128 | # Pyre type checker 129 | .pyre/ 130 | 131 | ./qwgc/utils/notification.py -------------------------------------------------------------------------------- /.travis.yml: -------------------------------------------------------------------------------- 1 | language: python 2 | 3 | dist: bionic 4 | 5 | python: 6 | - 3.6 7 | 8 | install: 9 | - pip install -e . 10 | 11 | script: 12 | # This repository has NOT had tests 13 | # but Travis CI wouldn't allow us not to write 14 | # `script` so we run this meaning less command to avoid CI errors. 15 | - echo "TODO" 16 | -------------------------------------------------------------------------------- /.vscode/settings.json: -------------------------------------------------------------------------------- 1 | { 2 | "python.linting.flake8Enabled": true, 3 | "python.linting.pycodestyleEnabled": false, 4 | "python.linting.pydocstyleEnabled": false, 5 | "python.linting.enabled": true 6 | } -------------------------------------------------------------------------------- /CITATION.md: -------------------------------------------------------------------------------- 1 | 2 | # Citations 3 | 4 | - Barr, Katie, Toby Fleming, and Viv Kendon. "Simulation methods for quantum walks on graphs applied to perfect state transfer and formal language recognition." CoSMoS 2013 (2013): 1. 5 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | Apache License 2 | Version 2.0, January 2004 3 | http://www.apache.org/licenses/ 4 | 5 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 6 | 7 | 1. Definitions. 8 | 9 | "License" shall mean the terms and conditions for use, reproduction, 10 | and distribution as defined by Sections 1 through 9 of this document. 11 | 12 | "Licensor" shall mean the copyright owner or entity authorized by 13 | the copyright owner that is granting the License. 14 | 15 | "Legal Entity" shall mean the union of the acting entity and all 16 | other entities that control, are controlled by, or are under common 17 | control with that entity. For the purposes of this definition, 18 | "control" means (i) the power, direct or indirect, to cause the 19 | direction or management of such entity, whether by contract or 20 | otherwise, or (ii) ownership of fifty percent (50%) or more of the 21 | outstanding shares, or (iii) beneficial ownership of such entity. 22 | 23 | "You" (or "Your") shall mean an individual or Legal Entity 24 | exercising permissions granted by this License. 25 | 26 | "Source" form shall mean the preferred form for making modifications, 27 | including but not limited to software source code, documentation 28 | source, and configuration files. 29 | 30 | "Object" form shall mean any form resulting from mechanical 31 | transformation or translation of a Source form, including but 32 | not limited to compiled object code, generated documentation, 33 | and conversions to other media types. 34 | 35 | "Work" shall mean the work of authorship, whether in Source or 36 | Object form, made available under the License, as indicated by a 37 | copyright notice that is included in or attached to the work 38 | (an example is provided in the Appendix below). 39 | 40 | "Derivative Works" shall mean any work, whether in Source or Object 41 | form, that is based on (or derived from) the Work and for which the 42 | editorial revisions, annotations, elaborations, or other modifications 43 | represent, as a whole, an original work of authorship. For the purposes 44 | of this License, Derivative Works shall not include works that remain 45 | separable from, or merely link (or bind by name) to the interfaces of, 46 | the Work and Derivative Works thereof. 47 | 48 | "Contribution" shall mean any work of authorship, including 49 | the original version of the Work and any modifications or additions 50 | to that Work or Derivative Works thereof, that is intentionally 51 | submitted to Licensor for inclusion in the Work by the copyright owner 52 | or by an individual or Legal Entity authorized to submit on behalf of 53 | the copyright owner. For the purposes of this definition, "submitted" 54 | means any form of electronic, verbal, or written communication sent 55 | to the Licensor or its representatives, including but not limited to 56 | communication on electronic mailing lists, source code control systems, 57 | and issue tracking systems that are managed by, or on behalf of, the 58 | Licensor for the purpose of discussing and improving the Work, but 59 | excluding communication that is conspicuously marked or otherwise 60 | designated in writing by the copyright owner as "Not a Contribution." 61 | 62 | "Contributor" shall mean Licensor and any individual or Legal Entity 63 | on behalf of whom a Contribution has been received by Licensor and 64 | subsequently incorporated within the Work. 65 | 66 | 2. Grant of Copyright License. Subject to the terms and conditions of 67 | this License, each Contributor hereby grants to You a perpetual, 68 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 69 | copyright license to reproduce, prepare Derivative Works of, 70 | publicly display, publicly perform, sublicense, and distribute the 71 | Work and such Derivative Works in Source or Object form. 72 | 73 | 3. Grant of Patent License. Subject to the terms and conditions of 74 | this License, each Contributor hereby grants to You a perpetual, 75 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 76 | (except as stated in this section) patent license to make, have made, 77 | use, offer to sell, sell, import, and otherwise transfer the Work, 78 | where such license applies only to those patent claims licensable 79 | by such Contributor that are necessarily infringed by their 80 | Contribution(s) alone or by combination of their Contribution(s) 81 | with the Work to which such Contribution(s) was submitted. If You 82 | institute patent litigation against any entity (including a 83 | cross-claim or counterclaim in a lawsuit) alleging that the Work 84 | or a Contribution incorporated within the Work constitutes direct 85 | or contributory patent infringement, then any patent licenses 86 | granted to You under this License for that Work shall terminate 87 | as of the date such litigation is filed. 88 | 89 | 4. Redistribution. You may reproduce and distribute copies of the 90 | Work or Derivative Works thereof in any medium, with or without 91 | modifications, and in Source or Object form, provided that You 92 | meet the following conditions: 93 | 94 | (a) You must give any other recipients of the Work or 95 | Derivative Works a copy of this License; and 96 | 97 | (b) You must cause any modified files to carry prominent notices 98 | stating that You changed the files; and 99 | 100 | (c) You must retain, in the Source form of any Derivative Works 101 | that You distribute, all copyright, patent, trademark, and 102 | attribution notices from the Source form of the Work, 103 | excluding those notices that do not pertain to any part of 104 | the Derivative Works; and 105 | 106 | (d) If the Work includes a "NOTICE" text file as part of its 107 | distribution, then any Derivative Works that You distribute must 108 | include a readable copy of the attribution notices contained 109 | within such NOTICE file, excluding those notices that do not 110 | pertain to any part of the Derivative Works, in at least one 111 | of the following places: within a NOTICE text file distributed 112 | as part of the Derivative Works; within the Source form or 113 | documentation, if provided along with the Derivative Works; or, 114 | within a display generated by the Derivative Works, if and 115 | wherever such third-party notices normally appear. The contents 116 | of the NOTICE file are for informational purposes only and 117 | do not modify the License. You may add Your own attribution 118 | notices within Derivative Works that You distribute, alongside 119 | or as an addendum to the NOTICE text from the Work, provided 120 | that such additional attribution notices cannot be construed 121 | as modifying the License. 122 | 123 | You may add Your own copyright statement to Your modifications and 124 | may provide additional or different license terms and conditions 125 | for use, reproduction, or distribution of Your modifications, or 126 | for any such Derivative Works as a whole, provided Your use, 127 | reproduction, and distribution of the Work otherwise complies with 128 | the conditions stated in this License. 129 | 130 | 5. Submission of Contributions. Unless You explicitly state otherwise, 131 | any Contribution intentionally submitted for inclusion in the Work 132 | by You to the Licensor shall be under the terms and conditions of 133 | this License, without any additional terms or conditions. 134 | Notwithstanding the above, nothing herein shall supersede or modify 135 | the terms of any separate license agreement you may have executed 136 | with Licensor regarding such Contributions. 137 | 138 | 6. Trademarks. This License does not grant permission to use the trade 139 | names, trademarks, service marks, or product names of the Licensor, 140 | except as required for reasonable and customary use in describing the 141 | origin of the Work and reproducing the content of the NOTICE file. 142 | 143 | 7. Disclaimer of Warranty. Unless required by applicable law or 144 | agreed to in writing, Licensor provides the Work (and each 145 | Contributor provides its Contributions) on an "AS IS" BASIS, 146 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or 147 | implied, including, without limitation, any warranties or conditions 148 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A 149 | PARTICULAR PURPOSE. You are solely responsible for determining the 150 | appropriateness of using or redistributing the Work and assume any 151 | risks associated with Your exercise of permissions under this License. 152 | 153 | 8. Limitation of Liability. In no event and under no legal theory, 154 | whether in tort (including negligence), contract, or otherwise, 155 | unless required by applicable law (such as deliberate and grossly 156 | negligent acts) or agreed to in writing, shall any Contributor be 157 | liable to You for damages, including any direct, indirect, special, 158 | incidental, or consequential damages of any character arising as a 159 | result of this License or out of the use or inability to use the 160 | Work (including but not limited to damages for loss of goodwill, 161 | work stoppage, computer failure or malfunction, or any and all 162 | other commercial damages or losses), even if such Contributor 163 | has been advised of the possibility of such damages. 164 | 165 | 9. Accepting Warranty or Additional Liability. While redistributing 166 | the Work or Derivative Works thereof, You may choose to offer, 167 | and charge a fee for, acceptance of support, warranty, indemnity, 168 | or other liability obligations and/or rights consistent with this 169 | License. However, in accepting such obligations, You may act only 170 | on Your own behalf and on Your sole responsibility, not on behalf 171 | of any other Contributor, and only if You agree to indemnify, 172 | defend, and hold each Contributor harmless for any liability 173 | incurred by, or claims asserted against, such Contributor by reason 174 | of your accepting any such warranty or additional liability. 175 | 176 | END OF TERMS AND CONDITIONS 177 | 178 | APPENDIX: How to apply the Apache License to your work. 179 | 180 | To apply the Apache License to your work, attach the following 181 | boilerplate notice, with the fields enclosed by brackets "[]" 182 | replaced with your own identifying information. (Don't include 183 | the brackets!) The text should be enclosed in the appropriate 184 | comment syntax for the file format. We also recommend that a 185 | file or class name and description of purpose be included on the 186 | same "printed page" as the copyright notice for easier 187 | identification within third-party archives. 188 | 189 | Copyright [2020] [Ryosuke Satoh] 190 | 191 | Licensed under the Apache License, Version 2.0 (the "License"); 192 | you may not use this file except in compliance with the License. 193 | You may obtain a copy of the License at 194 | 195 | http://www.apache.org/licenses/LICENSE-2.0 196 | 197 | Unless required by applicable law or agreed to in writing, software 198 | distributed under the License is distributed on an "AS IS" BASIS, 199 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 200 | See the License for the specific language governing permissions and 201 | limitations under the License. 202 | -------------------------------------------------------------------------------- /MANIFEST.in: -------------------------------------------------------------------------------- 1 | requirements.txt -------------------------------------------------------------------------------- /Pipfile: -------------------------------------------------------------------------------- 1 | [[source]] 2 | name = "pypi" 3 | url = "https://pypi.org/simple" 4 | verify_ssl = true 5 | 6 | [dev-packages] 7 | 8 | [packages] 9 | qiskit = "*" 10 | tqdm = "*" 11 | umap-learn = "*" 12 | grakel-dev = "==0.1a6" 13 | autopep8 = "*" 14 | seaborn = "*" 15 | toml = "*" 16 | slackweb = "*" 17 | grakel = "==0.1b7" 18 | cryptography = "*" 19 | 20 | [requires] 21 | python_version = "3.7" 22 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | 2 | # qwgc 3 | 4 | [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/Apache-2.0) 5 | 6 | 7 | 8 | qwgc stands for *Q*uantum *W*alk *G*raph *C*lassifier. 9 | This project aims to classify graph data with high accuracy. 10 | The key ideas of quantum machine learning and quantum walk are [here](https://qwqmlft.github.io/QuantumFrontier). 11 | 12 | 13 | ## How to install 14 | 15 | First, clone remote repository to your local. 16 | 17 | `git clone https://github.com/qwqmlf/qwgc.git` 18 | 19 | and then, 20 | 21 | ```zsh 22 | cd qwgc 23 | pip install -e . 24 | ``` 25 | 26 | If you don't have any errors, your build is success. 27 | 28 | ## How to use 29 | 30 | [experiments.toml](https://github.com/qwqmlf/qwgc/blob/master/qwgc/experiments.toml) is the configuration (Hyper parameter) for QWGC. 31 | 32 | Put hyper parameters and then run 33 | 34 | ```zsh 35 | cd qwgc 36 | python QWGC.py 37 | ``` 38 | 39 | (This simulation may take a long time to finish) 40 | 41 | ## Tutorials 42 | 43 | [Here](./notebook/tutorial.ipynb) is the tutorials how to use this (Japanese). 44 | 45 | ## Acknowledgement 46 | 47 | This project is supported by IPA mitou target project. 48 | -------------------------------------------------------------------------------- /Untitled.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": 2, 6 | "metadata": { 7 | "ExecuteTime": { 8 | "end_time": "2020-06-23T08:08:36.362262Z", 9 | "start_time": "2020-06-23T08:08:36.357553Z" 10 | } 11 | }, 12 | "outputs": [ 13 | { 14 | "data": { 15 | "text/plain": [ 16 | "['/Users/ryosuke/opt/anaconda3/lib/python3.7/site-packages/qiskit']" 17 | ] 18 | }, 19 | "execution_count": 2, 20 | "metadata": {}, 21 | "output_type": "execute_result" 22 | } 23 | ], 24 | "source": [ 25 | "import qiskit\n", 26 | "qiskit.__path__" 27 | ] 28 | }, 29 | { 30 | "cell_type": "code", 31 | "execution_count": null, 32 | "metadata": {}, 33 | "outputs": [], 34 | "source": [] 35 | } 36 | ], 37 | "metadata": { 38 | "jupytext": { 39 | "text_representation": { 40 | "extension": ".py", 41 | "format_name": "light", 42 | "format_version": "1.5", 43 | "jupytext_version": "1.3.3" 44 | } 45 | }, 46 | "kernelspec": { 47 | "display_name": "Python 3", 48 | "language": "python", 49 | "name": "python3" 50 | }, 51 | "language_info": { 52 | "codemirror_mode": { 53 | "name": "ipython", 54 | "version": 3 55 | }, 56 | "file_extension": ".py", 57 | "mimetype": "text/x-python", 58 | "name": "python", 59 | "nbconvert_exporter": "python", 60 | "pygments_lexer": "ipython3", 61 | "version": "3.7.4" 62 | }, 63 | "varInspector": { 64 | "cols": { 65 | "lenName": 16, 66 | "lenType": 16, 67 | "lenVar": 40 68 | }, 69 | "kernels_config": { 70 | "python": { 71 | "delete_cmd_postfix": "", 72 | "delete_cmd_prefix": "del ", 73 | "library": "var_list.py", 74 | "varRefreshCmd": "print(var_dic_list())" 75 | }, 76 | "r": { 77 | "delete_cmd_postfix": ") ", 78 | "delete_cmd_prefix": "rm(", 79 | "library": "var_list.r", 80 | "varRefreshCmd": "cat(var_dic_list()) " 81 | } 82 | }, 83 | "types_to_exclude": [ 84 | "module", 85 | "function", 86 | "builtin_function_or_method", 87 | "instance", 88 | "_Feature" 89 | ], 90 | "window_display": false 91 | } 92 | }, 93 | "nbformat": 4, 94 | "nbformat_minor": 4 95 | } 96 | -------------------------------------------------------------------------------- /Untitled.py: -------------------------------------------------------------------------------- 1 | # --- 2 | # jupyter: 3 | # jupytext: 4 | # text_representation: 5 | # extension: .py 6 | # format_name: light 7 | # format_version: '1.5' 8 | # jupytext_version: 1.3.3 9 | # kernelspec: 10 | # display_name: Python 3 11 | # language: python 12 | # name: python3 13 | # --- 14 | 15 | import qiskit 16 | qiskit.__path__ 17 | 18 | 19 | -------------------------------------------------------------------------------- /archive/qmnist.py: -------------------------------------------------------------------------------- 1 | """ 2 | This is archive version of previous qmnist project. 3 | """ 4 | from qiskit import QuantumCircuit, QuantumRegister, ClassicalRegister 5 | from qiskit import execute, Aer 6 | from numpy import pi 7 | from sklearn.datasets import load_digits 8 | from sklearn.model_selection import train_test_split 9 | from tqdm import trange 10 | 11 | import numpy as np 12 | import random 13 | import copy 14 | import umap.umap_ as umap 15 | 16 | THETA_MIN, THETA_MAX = -pi, pi 17 | BACKEND = Aer.get_backend('qasm_simulator') 18 | SHOTS = 2048 19 | 20 | 21 | class QMNIST: 22 | """ 23 | QMMIST is for quantum classifier for MNIST 24 | 25 | This 26 | """ 27 | def __init__(self, encoder, alpha=0.0001, 28 | n_particle=20, iteration=50, w=0.8, Cp=1.3, Cg=1.1): 29 | self.encoder = encoder 30 | self.alpha = alpha 31 | self.n_particle = n_particle 32 | self.iteration = iteration 33 | self.w = w 34 | self.Cp = Cp 35 | self.Cg = Cg 36 | 37 | def fit(self, x=None, y=None): 38 | dim = len(x[0]) 39 | particles = np.array([[random.uniform(THETA_MIN, THETA_MAX) 40 | for j in range(dim)] 41 | for n in range(self.n_particle)]) 42 | velocities = np.array([[0 for i in range(dim)] for n in range(self.n_particle)]) 43 | personal_best_pos = copy.copy(particles) 44 | personal_best_score = [self._get_error(x, y, theta) 45 | for theta in particles] 46 | 47 | best_particle = np.argmin(personal_best_score) 48 | grobal_best_pos = personal_best_pos[best_particle] 49 | 50 | errors = [] 51 | accs = [] 52 | print("Training start!") 53 | for t in trange(self.iteration): 54 | for n in range(self.n_particle): 55 | ran_p = random.uniform(0, 1) 56 | ran_g = random.uniform(0, 1) 57 | particles[n] = particles[n] + velocities[n] 58 | velocities[n] = (self.w*velocities[n] + 59 | self.Cp*ran_p*(personal_best_pos[n]-particles[n]) + 60 | self.Cg*ran_g*(grobal_best_pos-particles[n])) 61 | score = self._get_error(x, y, particles[n]) 62 | if score < personal_best_score[n]: 63 | personal_best_score[n] = score 64 | personal_best_pos[n] = particles[n] 65 | 66 | best_particle = np.argmin(personal_best_score) 67 | grobal_best_pos = personal_best_pos[best_particle] 68 | 69 | error = self._get_error(x, y, grobal_best_pos) 70 | acc = self._get_accuracy(x, y, grobal_best_pos) 71 | print(error, acc, grobal_best_pos) 72 | errors.append(error) 73 | accs.append(acc) 74 | converg = [errors, accs] 75 | return grobal_best_pos, converg 76 | 77 | def _get_error(self, x, y, theta): 78 | counts = self._get_counts(x, theta) 79 | errors = [] 80 | for ct, lb in zip(counts, y): 81 | emp_x = np.array([ct.get(b, 0)/SHOTS for b in self.encoder]) 82 | err = self._error_func(emp_x, np.array(y)) 83 | errors.append(err) 84 | return np.mean(errors) 85 | 86 | def _get_counts(self, x, theta): 87 | qcs = self._classifier(x, theta) 88 | job = execute(qcs, backend=BACKEND, shots=SHOTS) 89 | result = job.result() 90 | counts = [result.get_counts(qc) for qc in qcs] 91 | # print(counts) 92 | return counts 93 | 94 | @staticmethod 95 | def _error_func(x, bx, delta=1e-9): 96 | return -np.sum(bx * np.log(x + delta)) 97 | 98 | @staticmethod 99 | def _map_func(x): 100 | val = x/np.arcsinh(x) 101 | # print(val, x) 102 | return val 103 | 104 | def _classifier(self, x, theta): 105 | qcs = [] 106 | ld = len(x[0]) 107 | for xt in x: 108 | dataq = QuantumRegister(ld) 109 | c = ClassicalRegister(ld) 110 | qc = QuantumCircuit(dataq, c) 111 | for xd, qr in zip(xt, dataq): 112 | qc.h(qr) 113 | qc.rz(self._map_func(xd), qr) 114 | qc.h(qr) 115 | qc.rz(self._map_func(xd), qr) 116 | # anzatz 117 | for r in range(4): 118 | for ith, th in enumerate(theta): 119 | qc.ry(th, dataq[ith]) 120 | for ids, d in enumerate(dataq[:-1]): 121 | qc.cz(d, dataq[ids+1]) 122 | qc.cz(dataq[0], dataq[-1]) 123 | qc.measure(dataq, c) 124 | qcs.append(qc) 125 | return qcs 126 | 127 | def _get_accuracy(self, x, y, theta): 128 | counts = self._get_counts(x, theta) 129 | answers = self._get_answer(counts) 130 | count = 0 131 | for ans, lb in zip(answers, y): 132 | if ans == np.argmax(lb): 133 | count += 1 134 | accuracy = count/len(y) 135 | return accuracy 136 | 137 | def _get_answer(self, counts): 138 | answers = [] 139 | for cs in counts: 140 | answer = np.argmax([cs.get(b, 0) for b in self.encoder]) 141 | answers.append(answer) 142 | return answers 143 | 144 | def test(self, x, param): 145 | counts = self._get_counts(x, param) 146 | answer = self._get_answer(counts) 147 | return answer 148 | 149 | def performance(self, ans, label): 150 | count = 0 151 | for an, lb in zip(ans, label): 152 | if np.argmax(lb) == an: 153 | count += 1 154 | return count/len(label) 155 | 156 | 157 | if __name__ == '__main__': 158 | digits = load_digits() 159 | # qmnist = QMNIST(['0000000001', '0000000010', 160 | # '0000000100', '0000001000', 161 | # '0000010000', '0000100000', 162 | # '0001000000', '0010000000', 163 | # '0100000000', '1000000000']) 164 | 165 | qmnist = QMNIST(['0000', '1000', 166 | '0001', '1001', 167 | '0010', '1010', 168 | '0100', '1101', 169 | '0101', '1111']) 170 | 171 | # qmnist = QMNIST(['01', '10']) 172 | onhot_labels = [] 173 | for i in digits.target: 174 | lab = [0 for _ in range(10)] 175 | lab[int(i)] = 1 176 | onhot_labels.append(lab) 177 | reducer = umap.UMAP(n_components=4) 178 | reducer.fit(digits.data) 179 | embedding = reducer.transform(digits.data) 180 | # Verify that the result of calling transform is 181 | # idenitical to accessing the embedding_ attribute 182 | assert(np.all(embedding == reducer.embedding_)) 183 | x_train, x_test, y_train, y_test = train_test_split(embedding[:1000], onhot_labels[:1000]) 184 | param, conv = qmnist.fit(x_train, y_train) 185 | answers = qmnist.test(x_test, param) 186 | accs = qmnist.performance(answers, y_test) 187 | print(param, conv) 188 | print(accs) 189 | -------------------------------------------------------------------------------- /docs/DOCUMENT.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qwqmlf/qwgc/d06c805023bf37a1252505f7dc6e30461d861440/docs/DOCUMENT.md -------------------------------------------------------------------------------- /docs/document.txt: -------------------------------------------------------------------------------- 1 | 量子ウォークによるグラフ分類モデル 2 | 3 | 近年様々な場面においてグラフのデータの需要というものが高まってきています．例えば化学構造を判定する判定器や，ネットワークにおけるコミュニティ分析等様々に渡ります． 4 | 量子技術の関連で言うと，量子コンピュータにおける量子ビットをどのように接続するのか，といったものもグラフのデータの一種になります． 5 | しかしながらこのようなグラフのデータというのは一般に処理が非常に重たいという特徴があります．例えばグラフを行列の形式で表した場合，1000個のノードを持つグラフであれば 6 | 1000×1000の行列を扱うことになります．大きなグラフになれば，数万から数十万ものノードを持つグラフを扱わなければなりません．そのような場合古典のコンピュータでは非常に 7 | 大きなメモリが必要となってしまうことに加えて，処理に非常に時間がかかります．また，近年様々な機械学習の技術が発展してきている中で，グラフのデータにおける機械学習ではまだ 8 | 高い精度を実現しているとは言い難い状況です．特定のデータセットにおいては95%の精度を超えるものも出てきましたが，クラスやノードが増えたりすると50%を切ってしまうデータセットもあります． 9 | 10 | そこで今回我々が取り組んだこととしては，量子コンピュータを使ってうまくグラフを扱うことができないかということについて，実際にプロトタイプのモデルを作成し，検証を行いました． 11 | グラフに関するタスクは様々にありますが，今回はグラフの分類問題に絞ってモデルを作成しました． 12 | 13 | 要素技術について 14 | 量子ウォーク 15 | 16 | 量子ウォークとは簡単に言ってしまえば，古典のランダムウォークと呼ばれるものの量子版です．しかしながらその性質の特異さから，量子ランダムウォークではなく量子ウォーク 17 | と呼ばれるようになりました．量子ウォークは大きく二つの種類に分かれており，離散時間量子ウォークと連続時間量子ウォークと呼ばれるものが存在します．今回のプロジェクトでは 18 | コイン量子ウォークと呼ばれる離散時間量子ウォークの一種を用います． 19 | 20 | まずコイン量子ウォークの特徴として，古典のランダムウォークにおいて，各面の出現確率が1/2のコイントスをするように，コインのユニタリというものを定義します．この時コインは 21 | ユニタリな行列であればどのような行列でもコインとなり得ます．そしてそのようなコインに応じて，量子ウォーカーが次のステップでどのように動くのかということが決定されます． 22 | 古典のランダムウォークと全く異なる点のひとつとして，量子ウォーカーが確率的な場所の重ね合わせを持ちながら，歩を進めていくという点です．これにより，確率分布の拡散速度は 23 | 古典のランダムウォークに比べて平方根的に加速するという特徴があります．一般に一次元直線上のランダムウォークは無限回試行を繰り返せば，原点に確率分布が収束するような， 24 | 正規分布へ近づいていきますが，一次元直線状の量子ウォークにおいては，確率が外側に広がっていき，原点に向かって小さくなっていくような分布を示します．この分布の形状が 25 | どのようになるかについては，コインの決定方法によります． 26 | 27 | このような特徴を持った量子ウォークというものを任意のグラフ上で定義します．まず，グラフがどのように接続されているのかという隣接行列から，量子ウォーカーがどのように動くのか 28 | という情報を持ったステップオペレーターというものを定義します．各ノードにおいて，隣接するノードへどのように移動するのかを記述するための行列になります．次にどのくらいの 29 | 確率で移動するのかという確率遷移の役割を果たすコインオペレーターというものを定義します．今回は各ノードに対してローカルなコインオペレーターというものを定義します．これにより， 30 | 各ノード毎に隣接するノードのにどのくらいの確率で遷移するのかということを行列として保持しておくことができます．こうして，ステップオペレーターとコインオペレーターという二つの 31 | オペレーターが完成し，これらの内積をとったものが，1ステップの量子ウォークを表す時間発展のオペレーターとなります．これをt回かけることで，tステップ後の量子ウォークの時間発展の 32 | オペレーションを行うことができます． 33 | 34 | このようにして，グラフ上の量子ウォークを定義することで，あるグラフからそのグラフに特有な確率振幅というものを取り出すことができます．これをデータとして実際に分類を進めて 35 | いきます． 36 | 37 | 分類回路 38 | 39 | 上のように量子ウォークを用いてグラフから生成した確率振幅を，実際に量子ビットを用いて分類していきます. 40 | 41 | まず，量子ウォークの確率振幅を持った量子ビットをデータビットとし，そこに対して新たにマッピングビットとして量子ビットを追加します．次に，データビットからマッピングビット 42 | に対して，コントロールオペレーションを用いて情報をマッピングしていきます．この時，このコントロールオペレーションを特定の角度を持ったコントロールローテションオペレーション 43 | として行うことで，この回転の角度を調整していきます．この調整のプロセスについては後述します．最後にマッピングビットを測定し，事前に取り決めたエンコーダーにそって，その測定 44 | 結果がどのクラスに分類されるのかということを計算します．そしてその結果からどれくらいのコストが存在するのかということを計算し，先ほどの回転角度のパラメーターに対して 45 | フィードバックを行います． 46 | 47 | 48 | 49 | 50 | -------------------------------------------------------------------------------- /docs/index.html: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | This is qwgc 6 | 7 | 8 |

Welcom to quantum computing world!

9 | 10 | 11 | -------------------------------------------------------------------------------- /image/qwgc_logo.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qwqmlf/qwgc/d06c805023bf37a1252505f7dc6e30461d861440/image/qwgc_logo.png -------------------------------------------------------------------------------- /nohup.out: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qwqmlf/qwgc/d06c805023bf37a1252505f7dc6e30461d861440/nohup.out -------------------------------------------------------------------------------- /notebook/Plots.py: -------------------------------------------------------------------------------- 1 | # --- 2 | # jupyter: 3 | # jupytext: 4 | # text_representation: 5 | # extension: .py 6 | # format_name: light 7 | # format_version: '1.5' 8 | # jupytext_version: 1.3.3 9 | # kernelspec: 10 | # display_name: Python 3 11 | # language: python 12 | # name: python3 13 | # --- 14 | 15 | # # Convergence plot 16 | 17 | e1=[0.6286408520623922, 0.6289525825569178, 0.5412940012637213, 0.5399199822334315, 0.49756061144296837, 0.5386496534421491, 0.5110201664171742, 0.4785323556159, 0.4879992233852226, 0.49388898393628267, 0.4227883738901349, 0.47665407322957315, 0.49194537115508497, 0.4784480810071889, 0.5067394759922192, 0.4815508310174932, 0.4480058931682556, 0.5191713001759654, 0.5025600642057176, 0.47646282461622946, 0.4620775882390643, 0.5346632906032827, 0.4886131751452513, 0.5068477238076842, 0.4455982578380637, 0.5182468309511442, 0.5194051409981214, 0.5102047452714895, 0.4224846316777891, 0.4369363754766283, 0.47643414498658937, 0.5205363676955092, 0.5191792358482716, 0.43761845483213274, 0.4410680839960249, 0.5124071742830791, 0.5298468258800075, 0.5061098855038695, 0.48256532980818484, 0.4913228065366606, 0.4972612354925298, 0.4978721811109297, 0.4054224929939014, 0.4689590445321974, 0.47656007392483984, 0.4383041892763407, 0.44729094116756796, 0.4355191217281943, 0.39603253338464445] 18 | e2=[0.6445889256767168, 0.6535514888883678, 0.650516819829064, 0.5756429937727983, 0.6142765741620072, 0.5835705610683382, 0.5912615393329274, 0.6146589955098392, 0.6122305885006271, 0.615096288252815, 0.5755860024839088, 0.544477825409265, 0.5426761288272395, 0.5468697195293689, 0.564680589127484, 0.5771165305986262, 0.5684633735633585, 0.5680884419879598, 0.5612183060968275, 0.559276953305883, 0.43673818405890585, 0.4411540650165362, 0.4598893136551889, 0.43005624189179076, 0.45188026405543275, 0.47144659911395703, 0.5244689044281716, 0.41605587984112, 0.4556096201350907, 0.46408102111806254, 0.4693180656437913, 0.4268421245136888, 0.4412494204726097, 0.39367845982910615] 19 | e3=[0.662780905315912, 0.557526130143123, 0.5584640419156808, 0.5344490232122784, 0.593609022343057, 0.4937605123293818, 0.49679500580245917, 0.6381938409925619, 0.4969497574371143, 0.5490326065330788, 0.6127879707910444, 0.5089002055028583, 0.5504401226212559, 0.6142692709979238, 0.5721990951978606, 0.5018530856716944, 0.48248730823997454, 0.5393567930056656, 0.41791994591838755, 0.5305910774126542, 0.561185300298215, 0.595460105472618, 0.5115104804811111, 0.4495276525911773, 0.43500401066793454, 0.4161182946891324, 0.42475182877301493, 0.5073871378890902, 0.47263913226342963, 0.5149631515053271, 0.598002020559863, 0.41349915170939194, 0.47894825230795474, 0.5622578158963416, 0.5723435539368699, 0.4462502586322237, 0.4093627693814017, 0.5492960278816822, 0.510461265579312, 0.5808173512459651, 0.5597837733339822, 0.5640112157072309, 0.44698468041157, 0.5070970659950728, 0.48072844551751587, 0.5272173485178996, 0.5376622571610833, 0.5735053315795404, 0.4092035721307091, 0.5440903033836487, 0.5011121569382986, 0.5527630387469133, 0.4941897766576979, 0.470192162722345, 0.47418095121567044, 0.48084680432628457, 0.4239088964655666, 0.46940819486126134, 0.4537334210540194, 0.4864943090841813, 0.4687166558652696, 0.49614205816549944, 0.42874339930129546, 0.4479324170994075, 0.40787039901158556, 0.48786153290305956, 0.40963323764273557, 0.4352159063157325, 0.4888689022690961, 0.47162666268555464, 0.44671000444880116, 0.4967216183541344, 0.4012512434083865, 0.46064594635128436, 0.4049871203999046, 0.47170955324407465, 0.48706359188208187, 0.39931433341039707] 20 | e4=[0.631157721764274, 0.626369631668824, 0.6435168046798363, 0.5756890928745755, 0.6805477459512386, 0.5457392115779026, 0.6670013898692743, 0.571441632419972, 0.5711226503295723, 0.5043703755755192, 0.5133700170618634, 0.4890578230830625, 0.48377985095373055, 0.5401161872295855, 0.5231897730483787, 0.5224417136122825, 0.48197008488419696, 0.5430021431008605, 0.48585065976376307, 0.4947049468193294, 0.5128519566834271, 0.5111700349984294, 0.5021240973161041, 0.44473351723091004, 0.5387269935977932, 0.5000663192282513, 0.47768543276690545, 0.4804686186752253, 0.5298907808582913, 0.4752329625280757, 0.44839780034261145, 0.4701039454704951, 0.4918735308810149, 0.4728677770754973, 0.4576550851711339, 0.502450222085708, 0.5163739719069254, 0.5213008138551934, 0.4652131556745651, 0.49846654870033724, 0.5137510967641185, 0.5446460345322064, 0.47656732480459313, 0.4516336465509245, 0.5271248472630446, 0.4559450905739887, 0.49738101909129345, 0.5253791961349298, 0.5135150601261252, 0.44079823430873305, 0.5324834268850102, 0.5035609323350201, 0.5182742329465138, 0.5504791701670207, 0.531284676694358, 0.474503007577341, 0.5366327134778044, 0.5011612366374507, 0.5127348471786759, 0.5019671684236049, 0.44189963776688185, 0.46295316911159845, 0.5348905595183858, 0.5269112370262575, 0.5090272490103386, 0.47609522731817105, 0.491370758355211, 0.4135600347595616, 0.4818512230353636, 0.512062390185624, 0.47933046522657563, 0.4934857435730518, 0.4806171151139013, 0.4608250672563326, 0.41268520144053455, 0.4918494520803791, 0.46745210963432454, 0.48071324978066593, 0.5057204328529747, 0.4996037331029218, 0.5413217090462604, 0.4508761552721001, 0.47468589579687376, 0.5642027975890482, 0.41545508292293726, 0.48314552116470366, 0.3997297679560264] 21 | e5=[0.6682606140119535, 0.5778862814702397, 0.571284816447605, 0.5732217554886555, 0.576681243443929, 0.5696738503322155, 0.561619312862779, 0.5550863827708744, 0.5376039510952384, 0.5209829601641807, 0.5787276254893167, 0.5649896199922552, 0.5685287073392442, 0.4857622856463922, 0.5184760505457569, 0.5477220163966926, 0.5677443951422002, 0.546248508606815, 0.48857192683520445, 0.4844910863766792, 0.4503989103757882, 0.42963829512934754, 0.36462205360641897] 22 | e6=[0.6052190978338532, 0.6168189279264902, 0.5742753886203199, 0.5916751574417829, 0.6499174495644433, 0.5572021688155816, 0.5875677462728359, 0.5882764329669297, 0.42224941745738764, 0.46388965858198655, 0.46894811449133583, 0.5386172982477864, 0.445138792658842, 0.5659195198584153, 0.4280846453576261, 0.5081342706424578, 0.591777700219891, 0.4300014148912618, 0.44029194146143813, 0.4655751631352384, 0.5144871176051569, 0.4302597905571364, 0.47086406673852405, 0.42197278841497904, 0.5666806007720256, 0.4684396130328333, 0.46515501410751087, 0.5854247908909535, 0.46820092150107384, 0.4542753960527196, 0.5136825397873981, 0.42490230795806355, 0.5681653948495334, 0.4982418526291404, 0.5297805255166854, 0.5255641991721257, 0.46604946752584037, 0.42125778606643877, 0.5572774702893012, 0.4857469072821239, 0.46583437875598993, 0.49471480783108884, 0.4960882334145554, 0.464275426856948, 0.4227442887486467, 0.4736773611229292, 0.49004926134414617, 0.49931503682149486, 0.4988062267806028, 0.49268082475005154, 0.4488809454972674, 0.5174211905808321, 0.42898509257033995, 0.41935662991751776, 0.4804755068098414, 0.5517276889852001, 0.6108761441851374, 0.5106562598517737, 0.5707976053860792, 0.5431404102118769, 0.610686118237394, 0.5640443974179815, 0.45304879473737386, 0.5000248112364134, 0.4302734132332332, 0.48721656167805666, 0.5295429629785897, 0.46752394358529753, 0.6145431584915981, 0.50988218844425, 0.49915956989048965, 0.5278205896473646, 0.620852541372503, 0.4237926102741802, 0.42282174131809014, 0.4208938526055965, 0.487461524736154, 0.4753131817344044, 0.4822918616140902, 0.44156923779603763, 0.41871631240516527, 0.46848062331449036, 0.46980532334976366, 0.4709761803139777, 0.3976063267327049] 23 | e7=[0.6343177780040268, 0.619599304033062, 0.6242120198917819, 0.6031668480623487, 0.6402201060305929, 0.6231041583237609, 0.6293624185342378, 0.6458794486207945, 0.6827538772276097, 0.6720506480634977, 0.6899104720222323, 0.669870406811187, 0.6512371597807493, 0.6920056581481943, 0.5914131484170293, 0.6232179447484109, 0.5801354474941912, 0.6763159966305158, 0.56685784250218, 0.6637311529482484, 0.5847709695092046, 0.5633961429876598, 0.560108233954407, 0.52832327254978, 0.5283157800182223, 0.49202826266850064, 0.46519065544218025, 0.458531046376133, 0.48787502470038335, 0.4633186448926279, 0.5231878042415427, 0.5232549990667295, 0.45401437184614385, 0.4628743973973229, 0.46133926646432816, 0.4803737244518355, 0.4908663462940766, 0.47381789765142834, 0.46783045177687643, 0.46850374045145793, 0.4681717836248394, 0.4664232598009578, 0.475161204989716, 0.4698246364777594, 0.46279370428901806, 0.47086318835181673, 0.47130924184490053, 0.48004062896072147, 0.5380318909358003, 0.522459093653992, 0.5086512464786042, 0.4940124113230001, 0.5084629644805148, 0.49907221807047564, 0.5060149840853123, 0.5399783523243797, 0.5411928928687932, 0.5080996494638225, 0.49238446160879457, 0.5090379936341687, 0.4514971394109261, 0.5069246117398007, 0.5122056635240435, 0.5183315031961354, 0.5247657452974124, 0.5358356076290758, 0.5328220546034671, 0.5280624432744783, 0.5274743097746725, 0.4975618714767399, 0.49785887670110296, 0.5172247834730592, 0.5139599573220014, 0.5470670151810273, 0.5289727214355933, 0.5291905638247228, 0.4802756837124825, 0.5105438294031261, 0.4966151637696548, 0.4622230385770683, 0.5028894815058449, 0.5204636049862856, 0.46890755144793184, 0.536715875218253, 0.5287019172764743, 0.49212818638466355, 0.4978716350387547, 0.47170291861259944, 0.537535312754868, 0.5285083683899722, 0.5435672463104038, 0.5362293614488318, 0.48409470635156077, 0.45942009997388766, 0.5383618196483511, 0.5066198888774862, 0.5322328585016302, 0.5380757401024095, 0.5307318705952703, 0.49557513053245256, 0.540221765590457, 0.5319199897476646, 0.4425432496155992, 0.5424946438115121, 0.4967073166823288, 0.522464533354847, 0.48612498629064954, 0.5236114041318946, 0.5270520927976627, 0.5375392206453816, 0.49356780560525965, 0.5055856811585618, 0.5432119154861312, 0.5048427773577933, 0.45663313202496075, 0.45486778712936404, 0.44971790073868045, 0.5149780350294215, 0.5083240728859777, 0.44894944420421445, 0.5042145757006968, 0.48460426559739705, 0.4967351042896734, 0.5044661611141689, 0.43596825173934706, 0.5746881408014808, 0.5560117818174737, 0.48581143997140513, 0.4178848063830119, 0.4124531613792558, 0.41128625935257995, 0.5130877411153815, 0.5702506434152459, 0.5191608795298068, 0.40421862548733356, 0.49370921494417785, 0.47747102686452786, 0.4501159143626001, 0.42434966281242287, 0.5694928908915956, 0.4707648838018859, 0.46308786979076044, 0.5674826460753027, 0.41319744667993896, 0.43378581408673195, 0.5479508775304427, 0.42830800640058475, 0.5040567675034388, 0.5102932814824029, 0.5352098084517855, 0.4711907835610163, 0.4135569939522152, 0.5645434706974641, 0.39392153030740434] 24 | e8=[0.5913407969216123, 0.5968027045561548, 0.5807132242546393, 0.6045555424432444, 0.5790465745724817, 0.5146519574308367, 0.5386476154031689, 0.5457725039582826, 0.5544308346069118, 0.5435916500554991, 0.5640535382781875, 0.5419371097075426, 0.5441329113708785, 0.5580210026385325, 0.539490557893939, 0.5427861674156166, 0.5546224894721118, 0.5497072631021152, 0.5562272792102939, 0.5191799008595017, 0.5651000039656113, 0.5648645310545621, 0.5442534059819037, 0.5442239567027761, 0.5569325490735977, 0.5570671528470494, 0.5707478917193466, 0.5520669114088126, 0.5560904525558917, 0.536186225898877, 0.567669985618328, 0.5404920496327675, 0.551565557944254, 0.5425809403646311, 0.5557752017202766, 0.5535524873136072, 0.54511963019625, 0.5578999806184163, 0.5430200803197988, 0.5600545158472466, 0.5495828613081718, 0.554367532743426, 0.5492933831513049, 0.5571854750471325, 0.542362247911772, 0.5480834794851243, 0.5262099870228901, 0.48032788936449317, 0.4782151124259759, 0.49380543526049225, 0.5064807825285621, 0.4947974190672037, 0.4937050456318475, 0.49423348593034644, 0.4934758389741058, 0.5088370646749851, 0.4945296999162127, 0.4980231753465324, 0.4966765276873942, 0.4921028081643773, 0.515243532046272, 0.505921332289337, 0.5040872928584723, 0.4934447759396242, 0.5075192382266717, 0.5097597347103172, 0.4742844879083192, 0.5442708524819456, 0.4999348179754739, 0.461829856728611, 0.4705040366840693, 0.4903857708284842, 0.4924730661893619, 0.5134988021147889, 0.4923443029208326, 0.4727467433790539, 0.5225885499341563, 0.47025597023306087, 0.49101448323887475, 0.5430488000164548, 0.46579466193938523, 0.5151981816002086, 0.5429112242988399, 0.5038012282289347, 0.524662590969537, 0.4958443962133329, 0.4577390007625348, 0.5445978798228361, 0.502963558160195, 0.4894387668322627, 0.5440032976839626, 0.4819860916037589, 0.4654945753767531, 0.5152589293772628, 0.49937169712317414, 0.4864882370386736, 0.4507856824420945, 0.46654016177363283, 0.5097213545853928, 0.5404550468061052, 0.5419278186965883, 0.4495814173506067, 0.5153273272247367, 0.45098667264602854, 0.46487961539249983, 0.5234411744015217, 0.5336811462704812, 0.4455712071824399, 0.5365523056268828, 0.5121995525704492, 0.4736035794709903, 0.4473335804233988, 0.4906554048692217, 0.43056341428429684, 0.48750676622735406, 0.46975202791639503, 0.41985163165128453, 0.42365061155579714, 0.46134934441088365, 0.46918184061159246, 0.4432015336494499, 0.4720567531651249, 0.4683567090279011, 0.42967603770654866, 0.4236602081687182, 0.4135466892675132, 0.45815086391436777, 0.3967556123885781] 25 | e9=[0.6650768332261353, 0.5839188975158505, 0.5465120888927283, 0.5550907987683472, 0.567639764704636, 0.5085926678869052, 0.5275917470790742, 0.541221468537061, 0.5704477431835283, 0.5448885505140948, 0.5132874077181068, 0.5101701010115886, 0.5310406275923955, 0.5454947990050796, 0.5625857434190685, 0.5159442036807962, 0.5988034093157687, 0.5943047004640979, 0.551393955716915, 0.5233365342078535, 0.4752450983304961, 0.5050691325599479, 0.5235954093501491, 0.4552512305179317, 0.5831833398610338, 0.4731778439749536, 0.5687028273171127, 0.5889767650973098, 0.6064331743720209, 0.440816151178571, 0.448903119770252, 0.44343653750230294, 0.5001408741585447, 0.49118657617553213, 0.467956850971661, 0.4825464062978623, 0.5399849464121956, 0.47311507777648387, 0.4968562934355171, 0.46528754202033956, 0.5009424715411169, 0.4498238228474625, 0.5111711895467378, 0.46172956095965534, 0.48795037350222076, 0.49779900451624604, 0.41431572809038547, 0.5548793389362827, 0.5592370825347295, 0.47399601371736727, 0.5362121023064179, 0.5389262776700031, 0.48785242517507516, 0.48497470394201003, 0.48853758754813664, 0.452062814504457, 0.48515275520764484, 0.5178452313538177, 0.4370453782697105, 0.5143343530357027, 0.4733976407685126, 0.4377874336752083, 0.4225158612975633, 0.4079474669212051, 0.5525408255038543, 0.5199513551123233, 0.4448075381183994, 0.4889228086780745, 0.5444860707172519, 0.5482472432312546, 0.4715690031152096, 0.5860841514937343, 0.48233223561238736, 0.5260061348833885, 0.5273267432013465, 0.5210076022602085, 0.5551322110634548, 0.43377446518549584, 0.5930831673002215, 0.47734011874861937, 0.5223724082746285, 0.46364564600104363, 0.49103227014858114, 0.5078208924726832, 0.5142013832446156, 0.42539199248715004, 0.573131639889572, 0.4196562315745191, 0.5830005193887268, 0.5093072734962075, 0.5148966644389248, 0.4243085194017081, 0.6687310583399344, 0.46364319971652684, 0.49546870554391054, 0.5366528704484373, 0.5198103049962106, 0.5387543007216663, 0.582636034027522, 0.5363235009290618, 0.48327522215619056, 0.541926394132778, 0.5559785869462495, 0.6016761352697166, 0.5371624283020867, 0.5674319149578625, 0.5300704977548727, 0.5547608427601932, 0.4798895004992304, 0.4660034751436862, 0.524066221672807, 0.48431688551826535, 0.4902196506170733, 0.540522649117206, 0.4411247935696716, 0.5057168246330429, 0.6111460839056243, 0.5640026472940163, 0.5352351126005239, 0.43094285001997823, 0.5156524893462651, 0.43502378139170633, 0.4967973382630091, 0.49812092740202074, 0.5167071715045539, 0.5532030008701334, 0.5617041929433927, 0.5188911940730441, 0.5743757589835153, 0.5189928236553618, 0.4380557607907093, 0.5418787229350316, 0.44821135345496893, 0.5310716701733434, 0.5461678917843644, 0.519134739798097, 0.5511987988747173, 0.5773381630359999, 0.5233184425743422, 0.49873902095696, 0.5038420863478974, 0.5758701606630139, 0.5551576247842767, 0.5159133205832385, 0.5433872211476676, 0.5233134104986293, 0.5134074229814606, 0.49244212787856195, 0.5004507865528482, 0.5177767871179224, 0.5125121555910681, 0.5202937524719504, 0.5158894931633854, 0.518259284945597, 0.49879019957162246, 0.5138751753512852, 0.5193327327927922, 0.520053378362557, 0.4996759240244054, 0.5307492186079692, 0.5321791654099939, 0.512370357644313, 0.5051488492226954, 0.5155795339277843, 0.5850895438525353, 0.49333688234630446, 0.5146387574915534, 0.5152962059410358, 0.4916785527685588, 0.5220270697847215, 0.5108044898684504, 0.5225887670058841, 0.5054755192744912, 0.5200111921446353, 0.6186353617217828, 0.5162420347066328, 0.48991362080819656, 0.49458247546427553, 0.5310543800879368, 0.49156501831817023, 0.5209129827957445, 0.5071274285234982, 0.514889685151172, 0.4876652826616725, 0.5176711547960228, 0.501579488061985, 0.5219832016573605, 0.5211421406724956, 0.6102793042683985, 0.5092234796765647, 0.505841113259587, 0.5191361626567386, 0.509224562794439, 0.5095766326615782, 0.502079751965955, 0.5127003003234921, 0.5099580442929842, 0.5107503813960915, 0.5886776852182013, 0.5220688084678431, 0.5280708084273058, 0.5097898025489782, 0.5047902080968601, 0.5005626889772651, 0.596540291245381, 0.4996027048629127, 0.4934349312367303, 0.5409843332211969, 0.5017001340834972, 0.5094353513995228, 0.5170367406199069, 0.530769105638814, 0.5176185091437396, 0.5094759296400767, 0.5193568466416617, 0.4989014112392066, 0.5279480159795279, 0.5029211093859308, 0.5002122616570341, 0.5009459192487525, 0.5161268696467814, 0.5122714761320836, 0.4996834477594777, 0.5003828952897394, 0.5152310440552359, 0.5268164219017487, 0.4977712336603499, 0.5046063697037492, 0.5049611779603863, 0.471537858447829, 0.426280740636709, 0.4063048376824669, 0.4556689882653091, 0.429043458117716, 0.46098399335733065, 0.4783089903682395, 0.4548024144447215, 0.46235139284638144, 0.4855141050547884, 0.4525664843675803, 0.4752750869575304, 0.39445032834640115] 26 | e10=[0.6443296151657899, 0.5846714977547917, 0.5672715023778702, 0.5730989175581916, 0.5384647413211849, 0.645025012113611, 0.5818761835357733, 0.5485799801801237, 0.5470440562426591, 0.6848038874433044, 0.5762873974699276, 0.5210875247502859, 0.5840268965938661, 0.5996170002320559, 0.5675317667073588, 0.530705152494162, 0.43856855545184587, 0.5336486205176295, 0.5041957778265646, 0.5331599343293132, 0.4638759111952121, 0.5702342410880747, 0.4641523004965677, 0.49632076472769915, 0.4134526096867966, 0.6395538621427509, 0.5807024881555874, 0.54209316417008, 0.4917106519049478, 0.5746860150482541, 0.4903252002874721, 0.407509811691121, 0.41193636837385816, 0.5197750790448739, 0.5015629445346276, 0.5122720293420352, 0.5089435162867448, 0.479120258101935, 0.9612299362567398, 0.5322306367524896, 0.5408316328291428, 0.5005022853080526, 0.40517976682543727, 0.4738710242232649, 0.5853321607871663, 0.6350323074683683, 0.6220323108115962, 0.5756956691561051, 0.5785318277560356, 0.5495550621945828, 0.6706174816054492, 0.6379782769471595, 0.7984002265429345, 0.5819019958048154, 0.565515984004764, 0.585069528132984, 0.7163610180148806, 0.5907315843404257, 0.5649416080557663, 0.5797493081636891, 0.5929644452090546, 0.7590981885358427, 0.6029722793981109, 0.753962026409037, 0.6612429927176287, 0.5633060981382474, 0.6934616792472114, 0.599324234927781, 0.6621616927728853, 0.5921192185054662, 0.43641299280375273, 0.5122249502827196, 0.409546418406672, 0.42286517450873307, 0.6224070628071293, 0.5023519159195332, 0.5663255019772804, 0.5678506389375136, 0.464807663895237, 0.4602657015700848, 0.41183164602462163, 0.4275251302283884, 0.5801912671155873, 0.5318028256314792, 0.5397358271202957, 0.4531821739411755, 0.44849727924771193, 0.46869408154450576, 0.4549633610858585, 0.4462452859369011, 0.524381047614326, 0.379113122281514] 27 | errors = [e1, e2, e3, e4, e5, e6, e7, e8, e9] 28 | 29 | import numpy as np 30 | import matplotlib.pyplot as plt 31 | import seaborn as sns 32 | 33 | fig = plt.figure(figsize=(20, 10)) 34 | sns.set() 35 | plt.xlabel("The number of iterations", fontsize=30) 36 | plt.ylabel("Error", fontsize=30) 37 | plt.xticks(fontsize=20) 38 | plt.yticks(fontsize=20) 39 | for ier, er in enumerate(errors): 40 | lens = len(er) 41 | plt.plot(range(lens), er, label="ex%d"%(ier+1)) 42 | plt.legend(fontsize=20) 43 | plt.show() 44 | 45 | figure = plt.figure(figsize=(10, 10)) 46 | plt.boxplot([0.9473684210526315, 0.7894736842105263, 0.8947368421052632, 0.8947368421052632, 0.7894736842105263, 0.8947368421052632, 0.6842105263157895, 0.7368421052631579, 0.8888888888888888, 0.7777777777777778]) 47 | plt.title("Result of 10 cross validation", fontsize=20) 48 | plt.xlabel("MUTAG experiment", fontsize=20) 49 | plt.tick_params(labelsize=14) 50 | plt.ylabel("Accuracy", fontsize=20) 51 | 52 | e3s=[0.6114993145170406, 0.5580492397062562, 0.551723552407508, 0.5484116800260197, 0.5143462721841497, 0.628101523805724, 0.5356993571948219, 0.5266942280148578, 0.5141310873542365, 0.5638829554297106, 0.5496228055765812, 0.6360133689458043, 0.5201943196959011, 0.5481851459170456, 0.5055907141722972, 0.4941810064365272, 0.49340466605586575, 0.4634544459337397, 0.4611237576608274, 0.45213919591606666, 0.4534383746210488, 0.5054618981122072, 0.5165870609303926, 0.45209150635940387, 0.46805949776885464, 0.46960440853671903, 0.467819372880231, 0.4741484070294051, 0.4820804390230995, 0.4717719920228282, 0.5693892077112471, 0.5511274894354884, 0.5033156791541581, 0.5225321272437906, 0.5036599855410948, 0.51329297604302, 0.46424100042295025, 0.48893838154754043, 0.491828375814068, 0.5032194807511081, 0.4509417716028073, 0.4795241201436505, 0.5327343558753157, 0.4878793485411833, 0.4968022638355129, 0.47016307388363854, 0.5139168596002424, 0.44169284992918284, 0.4970774248606773, 0.5008123176407452, 0.5167832520763408, 0.5080038420660207, 0.48930759606351343, 0.513371961439122, 0.43260350311028717, 0.4378572126596461, 0.44645119576095627, 0.43974257495457336, 0.44501068910222275, 0.5046275199007141, 0.5248927422341266, 0.5057426950560957, 0.44435592756939385, 0.5028020254976295, 0.43205956750536756, 0.5095266443472911, 0.43189189288242835, 0.48790827272910336, 0.4615726181458669, 0.4708616965051463, 0.44032234603084713, 0.5179107099016528, 0.447553212974451, 0.4866584089091045, 0.5095162013165208, 0.5159928986646231, 0.5044830183103161, 0.4292861696424971, 0.5314628116720455, 0.43598839835680664, 0.4577829631040909, 0.4873913033076141, 0.4503809344508515, 0.484440023394841, 0.48888482964533436, 0.4920285335157382, 0.5073922782465428, 0.44670952906510897, 0.49607219758491355, 0.4308090376369893, 0.5268897603175041, 0.503559006011311, 0.5325591799558704, 0.5112110616324055, 0.4217329296463216, 0.4287457162429004, 0.5387852988261452, 0.5221160368806622, 0.509923902633072, 0.4411366299303317, 0.43803455040269307, 0.5335288104235335, 0.5272146852894581, 0.4602714866419109, 0.4465667728038471, 0.49895069098977923, 0.44450285295324765, 0.5251224372640607, 0.5278343169334901, 0.4596689338797286, 0.5217282828884132, 0.42935151668392935, 0.46706278369369864, 0.511648879842166, 0.49038583458744545, 0.4240684556714285, 0.5085615349599282, 0.4667558025833304, 0.5225293906551759, 0.4791700925332441, 0.42187666877056285, 0.4444577598049048, 0.43180975598120047, 0.47279635019359195, 0.40434165271953953, 0.4567439578687747, 0.4047810725738134, 0.45285436709795845, 0.5106699339221253, 0.41760528844233397, 0.463230639757571, 0.4583953066465225, 0.47033013236271226, 0.46012210399176057, 0.48140415708781703, 0.4471153121689473, 0.485398357745934, 0.4602156421568477, 0.48008157966420123, 0.4202314394133098, 0.504311984447005, 0.4655888166277228, 0.43114982313289985, 0.46496997393407696, 0.46041725990525456, 0.4662905080385762, 0.43186752985533855, 0.5113262759807295, 0.4719745720033306, 0.47791479563461675, 0.46799098060250977, 0.40620578584396677, 0.42314041607019615, 0.4162314065076259, 0.41579571768218204, 0.43814693535584737, 0.47413006835703997, 0.4974571568255316, 0.4935654224734384, 0.48136060672275627, 0.4550978367425495, 0.49706958170713433, 0.44513768595224995, 0.4253361121582329, 0.44381941515869333, 0.4761198169714065, 0.47667447892671105, 0.48046363042033574, 0.5199316709111657, 0.48927857372412814, 0.4808599995841265, 0.49799963296315697, 0.4492087033317719, 0.46986478238223167, 0.506700053418057, 0.4987938011003559, 0.4544414206143903, 0.5377744716063275, 0.5140073014745788, 0.42192876338556845, 0.4486045885520557, 0.42644917513144487, 0.4116325393736495, 0.4322134994562949, 0.476302541798784, 0.48675134068911835, 0.41824997807779385, 0.426746519901415, 0.46480259963728177, 0.5214455692010816, 0.41548303889992255, 0.42933136120094717, 0.4838030541852354, 0.494936691128753, 0.49186212498168014, 0.4217631685828114, 53 | 0.45602912462583534, 0.5125161938121239, 0.4457479372848052, 0.46618893650081916, 0.48881100807439615, 0.43768350802896466, 0.4739599720363637, 0.4955716129352989, 0.4804280162630706, 0.48163906724030736, 0.4143071427832678, 0.48475848400436833, 0.44899270858594775, 0.4838938111763643, 0.4848282167826762, 0.4838980545767503, 0.4866190119509327, 0.4717631934896995, 0.5165489359356199, 0.48360361419074716, 0.4804091649028527, 0.5097297759525212, 0.4571519002443824, 0.4452585896454598, 0.4414109845063244, 0.4235666003629125, 0.5153951286649423, 0.4883078008311837, 0.4346874565124854, 0.5005732966822127, 0.417852787993552, 0.4145857127347467, 0.48591110524336795, 0.4345903953415311, 0.41977222100784434, 0.5371283981310165, 0.47305277802964524, 0.44317952697867435, 0.4036964309526516, 0.4581214250541329, 0.5463341411652219, 0.46045127199889063, 0.4501169516101683, 0.4179825918218954, 0.49900783694877715, 0.4892906123586632, 0.42840660739898967, 0.47162708436149237, 0.4345798849491902, 0.47562520954776094, 0.4340927110566402, 0.4313336037367724, 0.4129143693247415, 0.5135537564781154, 0.47204003930970917, 0.4210640712529034, 0.4245605754426493, 0.4743745984824515, 0.48794937883439804, 0.4180700741879826, 0.4912515867736354, 0.4903159358869692, 0.49091273489139003, 0.4327263671175196, 0.5112962766853835, 0.418128526296058, 0.4832865396091334, 0.48082133074596606, 0.4832623410972604, 0.41270459153098055, 0.42109973525643424, 0.4998801209675613, 0.46762469008572893, 0.4799947903769364, 0.4736394640658181, 0.47211493605277344, 0.5287411401320841, 0.5200724040148723, 0.5066205296433474, 0.5127730183590961, 0.44369916589791547, 0.48456698587240804, 0.5051576392479146, 0.5081283632369991, 0.4193870952105006, 0.43548458423797165, 0.5146117075667941, 0.6496553944513833, 0.4143930795231045, 0.4289819520927289, 0.5080801555292406, 0.5131411994688787, 0.43696130146368667, 0.5427646826654766, 0.45345802627109194, 0.5433482910534887, 0.5143319170186181, 0.5097298427495505, 0.5136637813655547, 0.5106507420608952, 0.5548887569702587, 0.7065321055530301, 0.5168715814597589, 0.41794267534812685, 0.43090336510087845, 0.7212155749399769, 0.6165544559697002, 0.46033737106777256, 0.5089387495523169, 0.5253975539361965, 0.5115566116956278, 0.42320202401689294, 0.5246174570570632, 0.6281310330202612, 0.4751536510310882, 0.5950102991564058, 0.4978649425068354, 0.46610446928565485, 0.544534190394418, 0.6016255287956038, 0.5631491919128527, 0.6833236214330077, 0.4110880226407702, 0.4722339929634415, 0.5323749024552461, 0.4891842913332914, 0.6762690545068252, 0.41950833442907715, 0.4876659176727193, 0.6046630614307239, 0.4928000945510993, 0.553452844399436, 0.4511162775004234, 0.47151456349690807, 0.4909478739757076, 0.4918155366451409, 0.4431381515130087, 0.48011543632848297, 0.6012506969150045, 0.4602080508155012, 0.4857964362690456, 0.554271610426304, 0.41699549081779524, 0.4613389595659844, 0.413520621478622, 0.5796217614630786, 0.4849478885457273, 0.405352719966207, 0.4885767886098703, 0.5219683081108104, 0.4626909895475748, 0.5026622317544714, 0.5683235089403735, 0.40180703773062326, 0.41191760583198633, 0.39797130773487593] 54 | 55 | plt.plot(range(len(e3s)), e3s) 56 | 57 | 58 | -------------------------------------------------------------------------------- /notebook/qwgc_tutorial.py: -------------------------------------------------------------------------------- 1 | # -*- coding: utf-8 -*- 2 | # --- 3 | # jupyter: 4 | # jupytext: 5 | # text_representation: 6 | # extension: .py 7 | # format_name: light 8 | # format_version: '1.5' 9 | # jupytext_version: 1.3.3 10 | # kernelspec: 11 | # display_name: Python 3 12 | # language: python 13 | # name: python3 14 | # --- 15 | 16 | # # Quantum Walk Graph Classifier 17 | 18 | # 今回の成果物である量子ウォークにおけるグラフの分類器に関するチュートリアル 19 | 20 | # + 21 | import numpy as np 22 | import random 23 | import copy 24 | import sys 25 | import networkx as nx 26 | import matplotlib.pyplot as plt 27 | 28 | from numpy import pi 29 | from tqdm import trange 30 | from grakel import datasets, Graph 31 | from sklearn.model_selection import KFold 32 | 33 | sys.path.append("../") 34 | from qwgc.QWGC import QWGC 35 | # - 36 | # # 量子ウォーク 37 | 38 | # このプロジェクトにおける肝となる量子アルゴリズムである量子ウォークというものについてです。詳しくは我々の[プロジェクトページ](https://qwqmlf.github.io/QuantumFrontier/article)をご覧ください。今回は[MUTAG](https://rdrr.io/cran/QSARdata/man/mutagen.html)と呼ばれるデータセットを用いて、量子ウォークがグラフ上で行われるということがどういうことなのかということについて見ていきます。 39 | 40 | # MUTAGを取ってくる 41 | Data = datasets.fetch_dataset('MUTAG', verbose=False) 42 | data_x, data_y = np.array(Data.data), np.array(Data.target) 43 | 44 | # まずはMUTAGとはどのようなデータなのかという点について見ていきます。代表として先頭10データを可視化していきたいと思います。 45 | 46 | # visualization of data 47 | subtract = 0 48 | lens = [] 49 | for d, l in zip(data_x[:10], data_y[:10]): 50 | print(l) 51 | plt.figure(figsize=(10, 10)) 52 | G = nx.DiGraph() 53 | connection = d[0] 54 | nodesize = [(i+1)**800 for i in d[1].values()] 55 | edge_weight = d[2] 56 | lens.append(len([i for i in d[1].values()])) 57 | adjacency = Graph(connection).get_adjacency_matrix() 58 | nodes = np.array([str(i+1) for i, _ in enumerate(adjacency)]) 59 | edges = [] 60 | weight = [] 61 | for i, v in edge_weight.items(): 62 | ed = [str(st-subtract) for st in list(i)] 63 | ed.append(v+1) 64 | edges.append(tuple(ed)) 65 | subtract = max(d[1].keys()) 66 | G.add_nodes_from(nodes) 67 | G.add_weighted_edges_from(edges) 68 | pos = nx.kamada_kawai_layout(G) 69 | edge_labels = {(i, j): w['weight'] for i, j, w in G.edges(data=True)} 70 | nx.draw_networkx_edge_labels(G, pos, edge_labels=edge_labels) 71 | nx.draw_networkx(G, pos, with_labels=True, alpha=0.8) 72 | plt.show() 73 | 74 | # これらがMUTAGと呼ばれている変異原性があるのかないのかというデータセットになります。10から30程度のノード数を持ったグラフデータの集まりで、各ノードはある元素を表していて、各リンクは元素間の結合を表しています。各リンクの間についているラベルはその結合がどのような結合なのかということを表しています。ではここの上で量子ウォークを行うと、どのような状態がえられるのかということを見ていきたいと思います。 75 | 76 | # 今回のプログラムはこちらの論文を参考としています。[Barr, Katie, Toby Fleming, and Viv Kendon. "Simulation methods for quantum walks on graphs applied to perfect state transfer and formal language recognition." Proceedings of the 2013 workshop on complex systems modelling and simulation, Milan, Italy. 2013.] 77 | 78 | # まずは量子ウォーカーが1ステップ進んだ時を考えます。スタートは全て0ノードです。 79 | 80 | # + 81 | from qwgc.preprocess.qwalk import QuantumWalk 82 | # 対象のデータ 83 | data = data_x[0] 84 | label = data_y[0] 85 | # 隣接行列 86 | adjacency = Graph(data[0]).get_adjacency_matrix() 87 | count = np.count_nonzero(adjacency)//2 88 | # 量子ウォークのハイパーパラメータ 89 | step = 1 90 | # 次数が2の場合のコインのパラメータ(今回は簡単のために全てアダマールコイン) 91 | coin = np.kron(np.identity(count), 1/np.sqrt(2)*np.array([[1, 1], [1, -1]])) 92 | # 初期状態 (0からスタート) 93 | initial = None 94 | # 量子ウォーカーが測定される確率 95 | qwalk = QuantumWalk(initial, coin, adjacency) 96 | qwalk.n_steps(step) 97 | probs = qwalk.calc_probs() 98 | 99 | # 描画 100 | plt.figure(figsize=(10, 10)) 101 | G = nx.DiGraph() 102 | # ノードの大きさで確率を表す 103 | connection = data[0] 104 | nodesize = [(i+0.1)*800 for i in probs] 105 | edge_weight = data[2] 106 | nodes = np.array([str(i+1) for i, _ in enumerate(adjacency)]) 107 | edges = [] 108 | weight = [] 109 | for i, v in edge_weight.items(): 110 | ed = [str(st) for st in list(i)] 111 | ed.append(v+1) 112 | edges.append(tuple(ed)) 113 | subtract = max(data[1].keys()) 114 | G.add_nodes_from(nodes) 115 | G.add_weighted_edges_from(edges) 116 | pos = nx.spring_layout(G) 117 | edge_labels = {(i, j): w['weight'] for i, j, w in G.edges(data=True)} 118 | nx.draw_networkx_edge_labels(G, pos, edge_labels=edge_labels) 119 | nx.draw_networkx(G, pos, with_labels=True, alpha=0.8, node_size=nodesize) 120 | plt.show() 121 | 122 | 123 | # - 124 | 125 | # このように初期のノードから隣り合った二つのノードにおいて量子ウォーカーが観測されうるのがわかるかと思います。 126 | # もう少しステップを進めてみます。 127 | 128 | def draw(nodesize): 129 | plt.figure(figsize=(10, 10)) 130 | G = nx.DiGraph() 131 | # ノードの大きさで確率を表す 132 | connection = data[0] 133 | nodesize = [(i)*800 for i in probs] 134 | edge_weight = data[2] 135 | nodes = np.array([str(i+1) for i, _ in enumerate(adjacency)]) 136 | edges = [] 137 | weight = [] 138 | for i, v in edge_weight.items(): 139 | ed = [str(st) for st in list(i)] 140 | ed.append(v+1) 141 | edges.append(tuple(ed)) 142 | subtract = max(data[1].keys()) 143 | G.add_nodes_from(nodes) 144 | G.add_weighted_edges_from(edges) 145 | pos = nx.spring_layout(G) 146 | edge_labels = {(i, j): w['weight'] for i, j, w in G.edges(data=True)} 147 | nx.draw_networkx_edge_labels(G, pos, edge_labels=edge_labels) 148 | nx.draw_networkx(G, pos, with_labels=True, alpha=0.8, node_size=nodesize) 149 | plt.show() 150 | 151 | 152 | data = data_x[0] 153 | label = data_y[0] 154 | # 隣接行列 155 | adjacency = Graph(data[0]).get_adjacency_matrix() 156 | count = np.count_nonzero(adjacency)//2 157 | # 量子ウォークのハイパーパラメータ 158 | for step in range(1, 10): 159 | # 次数が2の場合のコインのパラメータ(今回は簡単のために全てアダマールコイン) 160 | coin = np.kron(np.identity(count), 1/np.sqrt(2)*np.array([[1, 1], [1, -1]])) 161 | # 初期状態 (0からスタート) 162 | initial = None 163 | # 量子ウォーカーが測定される確率 164 | qwalk = QuantumWalk(initial, coin, adjacency) 165 | qwalk.n_steps(step) 166 | probs = qwalk.calc_probs() 167 | draw(probs) 168 | 169 | # やや見辛いですが、少しずつ観測される確率が広がっていっていることがわかると思います。 170 | 171 | # # 分類回路 172 | 173 | # これらを実際に量子回路に流して分類を行なっていきます。 174 | 175 | from qwgc.preprocess.qwfilter import QWfilter 176 | from qiskit import QuantumCircuit, QuantumRegister, ClassicalRegister 177 | from qiskit import Aer, execute, transpile 178 | data = data_x[0] 179 | label = data_y[0] 180 | # filter 181 | u3param = [pi, 0, pi/2] 182 | # step 183 | step = 3 184 | # initial 重ね合わせを初期状態として利用 185 | initial = "super" 186 | qwfilter = QWfilter(u3param, step, initial) 187 | # 今回は測定をせずに、振幅をそのまま量子回路にマッピングを行います。 188 | amplitude = qwfilter.single_amplitude(data) 189 | 190 | # これにより特定のステップが終了した後の量子ウォークの確率振幅を取り出すことができました。これをqiskitのinitialize関数というものを用いて量子ビットに情報として入れていきます。またこの時$2^n$の大きさのベクトルである必要があることから0埋めを行います。 191 | 192 | la = len(amplitude) 193 | new_amplitude = list(amplitude) + [0 for i in range(64 - 38)] 194 | print(len(new_amplitude)) 195 | 196 | nq = 6 197 | #　量子レジスタの定義 198 | qr = QuantumRegister(nq, name="quantum walk") 199 | ancilla = QuantumRegister(2, name="ancilla") 200 | # 古典レジスタの定義 201 | cr = ClassicalRegister(2) 202 | # 量子回路の定義 203 | qc = QuantumCircuit(qr, ancilla, cr) 204 | qc.draw(output='mpl') 205 | 206 | # このように合計8量子ビットを用意しました。このうち上6つが量子ウォークのデータ（実際の実験ではこの部分が7個になっています。）、下2つが補助の量子ビットになっています。ここに量子ウォーク後の確率振幅を入れていきます。 207 | 208 | qc.initialize(new_amplitude, qr) 209 | 210 | # そして、パラメータを初期化し、補助量子ビットに制御Ryゲートを用いてマッピングを行なっていきます。 211 | 212 | # 回転角を初期化 213 | theta = [np.random.uniform(-pi, pi) for i in range(nq)] 214 | for ith, th in enumerate(theta): 215 | qc.cry(th, qr[ith], ancilla[ith%2]) 216 | qc.draw(output="mpl") 217 | 218 | # 最後に補助量子ビットを測定します。 219 | 220 | qc.measure(ancilla, cr) 221 | 222 | # この測定によって、このグラフがどのクラスに分類されるのかということを見ていきます。今回は01をクラス-1、（つまり変異原性の性質が陰性）10をクラスの1(性質が陽性)とします。 223 | 224 | backend = Aer.get_backend("qasm_simulator") 225 | shots = 1024 226 | job = execute(qc, backend=backend, shots=shots) 227 | counts = job.result().get_counts(qc) 228 | dinom = counts.get('01', 0) + counts.get('10', 0) + 1e-10 229 | print("クラス-1である確率:", counts.get('01', 0)/dinom, "クラス1である確率:", counts.get('10', 0)/dinom) 230 | if counts.get('01', 0)/dinom > counts.get('10', 0)/dinom: 231 | answer = -1 232 | else: 233 | answer = 1 234 | print("このグラフはクラス ", answer, "です.") 235 | 236 | # このような分類結果になります。これを実際のラベルと比較をしてみます。 237 | 238 | print("正解は", data_y[1], "です。") 239 | 240 | # これらを全てのデータに対して行い、エラーを計算し実際にパラメータのアップデートを行なっていきます。 241 | -------------------------------------------------------------------------------- /notebook/random_walk.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": 1, 6 | "metadata": { 7 | "ExecuteTime": { 8 | "end_time": "2020-02-07T05:47:10.548954Z", 9 | "start_time": "2020-02-07T05:47:10.062430Z" 10 | } 11 | }, 12 | "outputs": [], 13 | "source": [ 14 | "%matplotlib inline\n", 15 | "import matplotlib.pyplot as plt\n", 16 | "import numpy as np" 17 | ] 18 | }, 19 | { 20 | "cell_type": "code", 21 | "execution_count": 15, 22 | "metadata": { 23 | "ExecuteTime": { 24 | "end_time": "2020-02-07T06:01:38.608976Z", 25 | "start_time": "2020-02-07T06:01:38.596435Z" 26 | } 27 | }, 28 | "outputs": [ 29 | { 30 | "data": { 31 | "text/plain": [ 32 | "array([-0.01535351, -0.5319249 , -1.66412157, ..., -0.35529907,\n", 33 | " -0.44420375, -0.30539867])" 34 | ] 35 | }, 36 | "execution_count": 15, 37 | "metadata": {}, 38 | "output_type": "execute_result" 39 | } 40 | ], 41 | "source": [ 42 | "random_walk = np.random.randn(100000)\n", 43 | "random_walk" 44 | ] 45 | }, 46 | { 47 | "cell_type": "code", 48 | "execution_count": 16, 49 | "metadata": { 50 | "ExecuteTime": { 51 | "end_time": "2020-02-07T06:01:40.414733Z", 52 | "start_time": "2020-02-07T06:01:38.745700Z" 53 | }, 54 | "scrolled": false 55 | }, 56 | "outputs": [ 57 | { 58 | "data": { 59 | "text/plain": [ 60 | "(array([ 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n", 61 | " 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n", 62 | " 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n", 63 | " 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n", 64 | " 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n", 65 | " 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n", 66 | " 0., 0., 0., 0., 0., 0., 1., 1., 0., 0., 0.,\n", 67 | " 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n", 68 | " 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n", 69 | " 0., 0., 1., 1., 0., 0., 0., 1., 0., 0., 0.,\n", 70 | " 0., 0., 0., 0., 0., 0., 1., 0., 0., 0., 1.,\n", 71 | " 0., 0., 1., 0., 1., 0., 0., 0., 0., 0., 0.,\n", 72 | " 1., 1., 0., 1., 1., 2., 0., 0., 0., 0., 1.,\n", 73 | " 1., 0., 0., 1., 1., 1., 0., 2., 1., 0., 1.,\n", 74 | " 0., 1., 1., 2., 2., 1., 1., 3., 3., 1., 0.,\n", 75 | " 1., 2., 1., 3., 1., 2., 1., 1., 2., 1., 2.,\n", 76 | " 3., 4., 3., 0., 3., 0., 1., 1., 3., 1., 0.,\n", 77 | " 3., 4., 5., 3., 1., 3., 3., 5., 2., 5., 3.,\n", 78 | " 4., 5., 1., 3., 5., 4., 6., 7., 2., 7., 7.,\n", 79 | " 3., 4., 7., 5., 4., 4., 3., 5., 2., 6., 7.,\n", 80 | " 5., 3., 4., 7., 2., 5., 11., 8., 13., 6., 5.,\n", 81 | " 9., 10., 5., 9., 13., 8., 10., 8., 11., 12., 15.,\n", 82 | " 8., 12., 7., 7., 8., 12., 15., 10., 18., 17., 15.,\n", 83 | " 13., 18., 14., 10., 12., 18., 20., 23., 20., 21., 17.,\n", 84 | " 25., 21., 21., 21., 16., 26., 15., 23., 24., 23., 21.,\n", 85 | " 27., 26., 21., 23., 25., 31., 24., 23., 28., 35., 24.,\n", 86 | " 34., 36., 26., 30., 27., 31., 40., 29., 36., 37., 38.,\n", 87 | " 45., 33., 37., 44., 27., 48., 34., 46., 42., 46., 38.,\n", 88 | " 46., 53., 46., 63., 50., 64., 55., 62., 44., 55., 51.,\n", 89 | " 48., 69., 37., 55., 49., 49., 50., 75., 73., 65., 76.,\n", 90 | " 57., 68., 64., 57., 85., 70., 75., 71., 75., 79., 79.,\n", 91 | " 79., 85., 76., 68., 100., 88., 81., 83., 97., 110., 97.,\n", 92 | " 96., 102., 105., 93., 88., 110., 118., 85., 124., 106., 105.,\n", 93 | " 103., 113., 112., 99., 138., 127., 114., 121., 134., 145., 141.,\n", 94 | " 143., 130., 142., 147., 127., 139., 141., 146., 147., 159., 142.,\n", 95 | " 147., 138., 153., 152., 153., 153., 161., 167., 185., 164., 162.,\n", 96 | " 197., 201., 172., 173., 185., 187., 148., 167., 209., 189., 196.,\n", 97 | " 170., 202., 180., 186., 199., 201., 243., 204., 226., 233., 215.,\n", 98 | " 193., 205., 225., 200., 229., 243., 214., 246., 204., 232., 209.,\n", 99 | " 236., 233., 230., 245., 244., 270., 225., 237., 265., 249., 248.,\n", 100 | " 288., 251., 273., 238., 241., 300., 263., 283., 283., 297., 272.,\n", 101 | " 269., 263., 274., 303., 296., 297., 294., 290., 305., 292., 296.,\n", 102 | " 316., 296., 302., 314., 305., 316., 301., 313., 293., 315., 322.,\n", 103 | " 321., 326., 342., 332., 305., 317., 351., 354., 366., 340., 306.,\n", 104 | " 364., 329., 343., 350., 334., 351., 376., 351., 372., 333., 349.,\n", 105 | " 362., 334., 358., 345., 351., 371., 325., 361., 349., 356., 372.,\n", 106 | " 338., 364., 322., 372., 366., 337., 364., 385., 379., 384., 358.,\n", 107 | " 358., 368., 381., 382., 365., 384., 401., 366., 360., 383., 377.,\n", 108 | " 352., 383., 384., 384., 354., 368., 340., 354., 399., 403., 372.,\n", 109 | " 389., 375., 340., 381., 370., 338., 389., 382., 347., 400., 343.,\n", 110 | " 377., 370., 342., 355., 365., 358., 354., 346., 296., 343., 346.,\n", 111 | " 316., 364., 326., 371., 355., 334., 325., 352., 365., 348., 353.,\n", 112 | " 333., 347., 334., 345., 356., 347., 355., 360., 352., 317., 322.,\n", 113 | " 342., 333., 310., 329., 317., 308., 322., 339., 323., 305., 342.,\n", 114 | " 339., 283., 291., 318., 313., 266., 297., 285., 302., 288., 290.,\n", 115 | " 281., 290., 299., 288., 271., 252., 248., 275., 256., 293., 252.,\n", 116 | " 293., 298., 264., 283., 261., 229., 242., 259., 247., 250., 255.,\n", 117 | " 248., 240., 237., 245., 251., 241., 241., 249., 214., 224., 220.,\n", 118 | " 226., 238., 219., 235., 242., 222., 171., 214., 209., 228., 201.,\n", 119 | " 225., 211., 191., 180., 189., 184., 173., 175., 200., 155., 170.,\n", 120 | " 186., 169., 174., 185., 185., 156., 155., 159., 156., 172., 158.,\n", 121 | " 152., 161., 170., 140., 143., 151., 144., 140., 129., 130., 122.,\n", 122 | " 142., 135., 119., 114., 113., 121., 139., 135., 125., 124., 123.,\n", 123 | " 98., 116., 116., 130., 121., 110., 106., 96., 106., 104., 100.,\n", 124 | " 88., 84., 98., 87., 80., 103., 91., 90., 104., 110., 90.,\n", 125 | " 82., 93., 83., 91., 83., 90., 78., 84., 78., 78., 63.,\n", 126 | " 77., 65., 58., 68., 68., 59., 62., 56., 70., 46., 62.,\n", 127 | " 49., 48., 45., 50., 61., 57., 54., 49., 46., 58., 52.,\n", 128 | " 46., 49., 46., 48., 49., 53., 50., 46., 55., 44., 34.,\n", 129 | " 30., 35., 45., 32., 34., 41., 41., 33., 35., 27., 28.,\n", 130 | " 32., 20., 33., 35., 22., 26., 29., 32., 20., 22., 21.,\n", 131 | " 30., 28., 20., 26., 19., 23., 22., 24., 24., 11., 29.,\n", 132 | " 17., 23., 23., 14., 17., 9., 16., 19., 15., 12., 19.,\n", 133 | " 14., 12., 18., 13., 16., 17., 13., 17., 18., 5., 10.,\n", 134 | " 5., 14., 9., 11., 10., 10., 11., 9., 9., 6., 10.,\n", 135 | " 10., 7., 5., 4., 9., 2., 4., 6., 7., 2., 3.,\n", 136 | " 9., 10., 7., 6., 10., 5., 9., 4., 7., 8., 3.,\n", 137 | " 4., 6., 5., 5., 4., 4., 6., 4., 6., 5., 2.,\n", 138 | " 4., 3., 4., 1., 2., 3., 4., 3., 0., 1., 2.,\n", 139 | " 4., 3., 1., 2., 5., 3., 5., 4., 0., 3., 2.,\n", 140 | " 0., 1., 1., 1., 2., 3., 4., 0., 2., 1., 0.,\n", 141 | " 2., 1., 0., 0., 1., 1., 0., 0., 4., 2., 2.,\n", 142 | " 1., 1., 1., 2., 1., 1., 1., 0., 0., 3., 0.,\n", 143 | " 0., 0., 3., 2., 0., 0., 2., 2., 0., 1., 1.,\n", 144 | " 1., 2., 2., 0., 2., 1., 0., 0., 0., 0., 0.,\n", 145 | " 0., 0., 0., 1., 0., 1., 0., 0., 0., 1., 0.,\n", 146 | " 0., 1., 0., 0., 0., 0., 0., 1., 0., 0., 1.,\n", 147 | " 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n", 148 | " 0., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n", 149 | " 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n", 150 | " 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.]),\n", 151 | " array([-4.89790212, -4.88865781, -4.87941351, ..., 4.32791098,\n", 152 | " 4.33715528, 4.34639958]),\n", 153 | " )" 154 | ] 155 | }, 156 | "execution_count": 16, 157 | "metadata": {}, 158 | "output_type": "execute_result" 159 | }, 160 | { 161 | "data": { 162 | "image/png": "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\n", 163 | "text/plain": [ 164 | "
" 165 | ] 166 | }, 167 | "metadata": { 168 | "needs_background": "light" 169 | }, 170 | "output_type": "display_data" 171 | } 172 | ], 173 | "source": [ 174 | "fig = plt.figure(figsize=(10, 10))\n", 175 | "plt.title(\"random walk\", fontsize=20)\n", 176 | "plt.xlabel(\"Position\", fontsize=20)\n", 177 | "plt.ylabel(\"Probability\", fontsize=20)\n", 178 | "plt.hist(random_walk, 1000)" 179 | ] 180 | }, 181 | { 182 | "cell_type": "code", 183 | "execution_count": null, 184 | "metadata": {}, 185 | "outputs": [], 186 | "source": [] 187 | } 188 | ], 189 | "metadata": { 190 | "jupytext": { 191 | "text_representation": { 192 | "extension": ".py", 193 | "format_name": "light", 194 | "format_version": "1.5", 195 | "jupytext_version": "1.3.3" 196 | } 197 | }, 198 | "kernelspec": { 199 | "display_name": "Python 3", 200 | "language": "python", 201 | "name": "python3" 202 | }, 203 | "varInspector": { 204 | "cols": { 205 | "lenName": 16, 206 | "lenType": 16, 207 | "lenVar": 40 208 | }, 209 | "kernels_config": { 210 | "python": { 211 | "delete_cmd_postfix": "", 212 | "delete_cmd_prefix": "del ", 213 | "library": "var_list.py", 214 | "varRefreshCmd": "print(var_dic_list())" 215 | }, 216 | "r": { 217 | "delete_cmd_postfix": ") ", 218 | "delete_cmd_prefix": "rm(", 219 | "library": "var_list.r", 220 | "varRefreshCmd": "cat(var_dic_list()) " 221 | } 222 | }, 223 | "types_to_exclude": [ 224 | "module", 225 | "function", 226 | "builtin_function_or_method", 227 | "instance", 228 | "_Feature" 229 | ], 230 | "window_display": false 231 | } 232 | }, 233 | "nbformat": 4, 234 | "nbformat_minor": 4 235 | } 236 | -------------------------------------------------------------------------------- /notebook/random_walk.py: -------------------------------------------------------------------------------- 1 | # --- 2 | # jupyter: 3 | # jupytext: 4 | # text_representation: 5 | # extension: .py 6 | # format_name: light 7 | # format_version: '1.5' 8 | # jupytext_version: 1.3.3 9 | # kernelspec: 10 | # display_name: Python 3 11 | # language: python 12 | # name: python3 13 | # --- 14 | 15 | # %matplotlib inline 16 | import matplotlib.pyplot as plt 17 | import numpy as np 18 | 19 | random_walk = np.random.randn(100000) 20 | random_walk 21 | 22 | fig = plt.figure(figsize=(10, 10)) 23 | plt.title("random walk", fontsize=20) 24 | plt.xlabel("Position", fontsize=20) 25 | plt.ylabel("Probability", fontsize=20) 26 | plt.hist(random_walk, 1000) 27 | 28 | 29 | -------------------------------------------------------------------------------- /qwgc/QWGC.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import random 3 | import copy 4 | import time 5 | 6 | from numpy import pi 7 | from tqdm import trange 8 | from grakel import datasets 9 | from sklearn.model_selection import KFold 10 | 11 | try: 12 | # FIXME 13 | from .classifier.qcircuit import ClassifierCircuit 14 | from .preprocess.qwfilter import QWfilter 15 | except ImportError: 16 | from classifier.qcircuit import ClassifierCircuit 17 | from preprocess.qwfilter import QWfilter 18 | 19 | try: 20 | from utils.notification import Notify 21 | notify = True 22 | except Exception: 23 | notify = False 24 | np.set_printoptions(threshold=10000) 25 | 26 | THETA_MIN, THETA_MAX = -pi, pi 27 | 28 | 29 | class QWGC: 30 | ''' 31 | In this package, we selected particle swarm optimzier as a optimizer 32 | for tuning vector theta. 33 | You can choose any of optimization way, but gradient way might not 34 | suit for this model. 35 | FIXME: more customizable 36 | In this scheme, the type of coin is constantly changing in each iteraions. 37 | ''' 38 | 39 | def __init__(self, encoder, Cp=1.8, Cg=1.3, n_particle=20, T=100, w=0.8, 40 | ro_max=1.0, n_layer=2, lamb=0.005, n_steps=5, 41 | initial='super', **kwargs): 42 | ''' 43 | Hyper parameters of model. 44 | Input: 45 | FIXME: 46 | encoder: list of binary 47 | e.g. ['01', '10'] 48 | length of encoder must be the same as the number of class 49 | Cp: float (constant value of coefficient 50 | for personal best position) 51 | Cg: float (constant value of coefficient 52 | for grobal best position) 53 | n_particle: int (the number of particles 54 | for searching better parameters) 55 | T: int (the number of iterations of how many times 56 | this model learns) 57 | w: float (constant value of coefficient 58 | for previous particle directions) 59 | ro_max: float (maximum number of random number 60 | which is used for updating parameter) 61 | n_layer: int (the number of layers of mapping circuit) 62 | lamb: float (constant value of coefficient 63 | for the sum of square in error) 64 | n_steps: int (the number of steps of Quantum walk) 65 | ''' 66 | if len(encoder[0]) != n_layer: 67 | raise ValueError('The size of encoder is different\ 68 | from the number of layers') 69 | self.encoder = encoder 70 | self.Cp = Cp 71 | self.Cg = Cg 72 | self.w = w 73 | self.ro_max = ro_max 74 | # FIXME more efficient 75 | if T < 0: 76 | raise ValueError('The number of iterations must be \ 77 | non negative value') 78 | if n_particle < 0: 79 | raise ValueError('The number of particles must be\ 80 | non negative value') 81 | if n_layer <= 0: 82 | raise ValueError('The number of layers must be\ 83 | one or over') 84 | if n_steps < 0: 85 | raise ValueError('The number of steps must be\ 86 | zero or over') 87 | self.T = T 88 | self.n_particle = n_particle 89 | self.layers = n_layer 90 | self.step = n_steps 91 | self.lamb = lamb 92 | self.initial = initial 93 | 94 | def optimize(self, train_data, train_label, coin_update=False): 95 | ''' 96 | Input: 97 | train_data: 2dim array (a series of training data) 98 | train_label: 2dim array (a series of label, one hot) 99 | Output: 100 | theta: array 101 | coin_param: ? 102 | ''' 103 | self.n_class = len(train_label[0]) 104 | if self.n_class > 2**self.layers: 105 | raise ValueError('the number of class must be less than 2^layers') 106 | 107 | # initial parameter for Quantum Walk 108 | # each particle has [theta, phi, lambda] 109 | # coin_u3s = np.array([[random.uniform(THETA_MIN, THETA_MAX) 110 | # for i in range(3)] 111 | # for n in range(self.n_particle)]) 112 | if coin_update: 113 | coin_u3s = np.array([[random.uniform(THETA_MIN, THETA_MAX), 114 | 0, 0] 115 | for n in range(self.n_particle)]) 116 | else: 117 | coin_u3s = np.array([[pi/2, pi/2, 0] 118 | for n in range(self.n_particle)]) 119 | 120 | ampdata = [QWfilter(coin_u3s[n], self.step, 121 | self.initial).amplitude(train_data) 122 | for n in range(self.n_particle)] 123 | theta_size = self.ceilog(max(set(map(len, ampdata[0])))) 124 | 125 | # FIXME 126 | n_amp = np.array([[self._zero_fill(amp, 2**theta_size) 127 | for amp in ampdata[n]] 128 | for n in range(self.n_particle)]) 129 | 130 | # initialize particles 131 | particles = np.array([np.array([random.uniform(THETA_MIN, THETA_MAX) 132 | for i in range(theta_size)]) 133 | for n in range(self.n_particle)]) 134 | 135 | velocities = np.array([np.zeros(theta_size) 136 | for n in range(self.n_particle)]) 137 | coin_v = np.array([np.zeros(3) for n in range(self.n_particle)]) 138 | 139 | # for recording best positions of each particle 140 | personal_bpos = copy.copy(particles) 141 | personal_cbpos = copy.copy(coin_u3s) 142 | personal_best_scores = [self._get_cost(amps, train_label, theta, notify=False) 143 | for amps, theta in zip(n_amp, particles)] 144 | 145 | # recording best position of all particles 146 | best_particle = np.argmin(personal_best_scores) 147 | grobal_best_pos = personal_bpos[best_particle] 148 | grobal_best_coin = coin_u3s[best_particle] 149 | # print(train_label) 150 | 151 | print('training start!') 152 | # start training 153 | # to check convergence of error, prepare this list 154 | errors = [] 155 | accuracy = [] 156 | flag = 1 157 | for nap in n_amp[best_particle]: 158 | print(list(nap)) 159 | for t in trange(self.T, desc='training'): 160 | ampdata = [QWfilter(coin_u3s[n], self.step, 161 | self.initial).amplitude(train_data) 162 | for n in range(self.n_particle)] 163 | n_amp = np.array([[self._zero_fill(amp, 2**theta_size) 164 | for amp in ampdata[n]] 165 | for n in range(self.n_particle)]) 166 | for n in range(self.n_particle): 167 | amp = n_amp[n] 168 | # random number for personal best pos 169 | rnp = random.uniform(0, self.ro_max) 170 | # random number for grobal best 171 | rng = random.uniform(0, self.ro_max) 172 | rnp_c = random.uniform(0, self.ro_max) 173 | rng_c = random.uniform(0, self.ro_max) 174 | 175 | # update positions parameter 176 | particles[n] = particles[n] + velocities[n] 177 | velocities[n] = (self.w*velocities[n] + 178 | self.Cp*rnp*(personal_bpos[n]-particles[n]) + 179 | self.Cg*rng*(grobal_best_pos-particles[n])) 180 | # update coins 181 | # coin_u3s[n] = coin_u3s[n] + coin_v[n] 182 | # No coin update 183 | if coin_update: 184 | coin_u3s[n] = coin_u3s[n] + coin_v[n] 185 | 186 | coin_v[n] = (self.w*coin_v[n] + 187 | self.Cp*rnp_c*(personal_cbpos[n]-coin_u3s[n]) + 188 | self.Cg*rng_c*(grobal_best_coin-coin_u3s[n])) 189 | 190 | # calculation cost with updated parameters 191 | # and update best position and score 192 | score = self._get_cost(amp, train_label, particles[n]) 193 | if score < personal_best_scores[n]: 194 | personal_best_scores[n] = score 195 | personal_bpos[n] = particles[n] 196 | personal_cbpos[n] = coin_u3s[n] 197 | 198 | # in all particles, calculate which is the best particle 199 | # and coin parameters 200 | best_particle = np.argmin(personal_best_scores) 201 | grobal_best_coin = coin_u3s[best_particle] 202 | grobal_best_pos = personal_bpos[best_particle] 203 | 204 | best_amp_data = QWfilter(grobal_best_coin, self.step, 205 | self.initial).amplitude(train_data) 206 | n_best_amp = [self._zero_fill(amp, 2**theta_size) 207 | for amp in best_amp_data] 208 | if t % 10 == 0: 209 | error = self._get_cost(n_best_amp, train_label, grobal_best_pos, notify=False) 210 | else: 211 | error = self._get_cost(n_best_amp, train_label, grobal_best_pos, notify=False) 212 | accs = self._get_accuracy(n_best_amp, train_label, 213 | grobal_best_pos) 214 | errors.append(error) 215 | accuracy.append(accs) 216 | # print(error, accs) 217 | # if t % 10 == 0: 218 | # flag *= -1 219 | # if notify: 220 | # Notify.notify_error(t, error, accs) 221 | if error < 0.40: 222 | break 223 | # if t > 10 and np.mean(errors[-10:-1]) < errors[-1]: 224 | # # print(particles) 225 | # reseter = np.array([[random.uniform(-pi/8, pi/8) for _ in range(theta_size)]for n in range(particles)]) 226 | # # print('reseter', reseter, 'best particle', best_particle) 227 | # particles = np.array([p+reseter for i, p in enumerate(particles) if i != best_particle]) 228 | # # print(particles) 229 | for nap in n_amp[best_particle]: 230 | print(list(nap)) 231 | convergence = [errors, accuracy] 232 | return grobal_best_pos, grobal_best_coin, convergence 233 | 234 | def _get_cost(self, data, label, theta, notify=False): 235 | cost = ClassifierCircuit(data, label, theta, self.n_class, 236 | self.layers, self.encoder).cost(notify=notify) 237 | 238 | error = cost + self.lamb*np.sum([i**2 for i in theta]) 239 | return error 240 | 241 | def _get_accuracy(self, data, label, theta): 242 | answers = ClassifierCircuit(data, label, theta, self.n_class, 243 | self.layers, self.encoder).answers() 244 | acc = self._accs(answers, label) 245 | return acc 246 | 247 | @staticmethod 248 | def ceilog(x): 249 | return int(np.ceil(np.log2(x))) 250 | 251 | def test(self, test_data, theta, coin_param): 252 | ''' 253 | Test function to evaluate the performance 254 | Input: 255 | test_data: 2dim array (a series of test data) 256 | theta: tuned parameters theta 257 | coin_param: optimzed coin parameters 258 | Output: 259 | answers: 2dim array (answers for each data) 260 | ''' 261 | theta_size = len(theta) 262 | ampdata = QWfilter(coin_param, self.step, 263 | self.initial).amplitude(test_data) 264 | n_amp = np.array([self._zero_fill(amp, 2**theta_size) 265 | for amp in ampdata]) 266 | # TODO Inplement the case that the number of class is unknown 267 | answers = ClassifierCircuit(n_amp, None, theta, 2, 268 | self.layers, self.encoder).answers() 269 | return answers 270 | 271 | def _accs(self, ans, label): 272 | count = 0 273 | for i, j in zip(ans, label): 274 | if np.argmax(i) == np.argmax(j): 275 | count += 1 276 | # print('answer ', [np.argmax(i) for i in ans]) 277 | # print('label ', [np.argmax(i) for i in label]) 278 | return count/len(label) 279 | 280 | @staticmethod 281 | def _zero_fill(x, base, array=True): 282 | 283 | # FIXME efficiently 284 | xl = list(x) 285 | x_len = len(xl) 286 | if base - x_len < 0: 287 | raise ValueError('Error') 288 | xs = xl + [0 for _ in range(base-x_len)] 289 | if array: 290 | return np.array(xs) 291 | else: 292 | return xs 293 | 294 | def summary(self, answer, label, printer=False): 295 | accs = self._accs(answer, label) 296 | if printer: 297 | print('The accuracy of this model is ', accs) 298 | return accs 299 | 300 | 301 | def one_hot_encoder(label, n_class): 302 | enc_label = [np.zeros(n_class) for _ in label] 303 | for ilb, lb in enumerate(label): 304 | if lb == -1: 305 | enc_label[ilb][0] = 1 306 | else: 307 | enc_label[ilb][lb-1] = 1 308 | return enc_label 309 | 310 | 311 | if __name__ == '__main__': 312 | # prepare dataset 313 | import toml 314 | # parsing parameters from toml 315 | config = toml.load('experiments.toml') 316 | p_pso = config['pso'] 317 | p_qw = config['qw'] 318 | 319 | data_name = 'ENZYMES' 320 | Data = datasets.fetch_dataset(data_name, verbose=False) 321 | data_x, data_y = np.array(Data.data), np.array(Data.target) 322 | 323 | acclist = [] 324 | k = 10 325 | kf = KFold(n_splits=k, shuffle=True, random_state=1) 326 | 327 | qwgc = QWGC(['000001', '000010', '000100', '001000', '010000', '100000'], 328 | Cp=p_pso['Cp'], Cg=p_pso['Cg'], 329 | n_particle=p_pso['particles'], T=p_pso['iterations'], 330 | w=p_pso['w'], ro_max=p_pso['random_max'], 331 | n_layer=p_pso['layers'], lamb=p_pso['lambda'], 332 | n_steps=p_qw['steps'], initial=p_qw['initial']) 333 | 334 | for train_index, test_index in kf.split(data_x): 335 | # preprocessing for generating data. 336 | x_train, y_train = data_x[train_index], data_y[train_index] 337 | x_test, y_test = data_x[test_index], data_y[test_index] 338 | # Notify.notify_accs("class%d" % y_train[0], "class%d" % y_train[-1]) 339 | 340 | # one hot encoding 341 | print(list(y_train)) 342 | y_train = one_hot_encoder(y_train, 6) 343 | y_test = one_hot_encoder(y_test, 6) 344 | 345 | theta, coin_param, conv = qwgc.optimize(x_train, y_train) 346 | # test 347 | ans = qwgc.test(x_test, theta, coin_param) 348 | # evaluate 349 | accs = qwgc.summary(ans, y_test) 350 | 351 | acclist.append(accs) 352 | print(accs) 353 | if notify: 354 | Notify.notify_accs(accs, conv) 355 | print('acclist', acclist) 356 | print('mean', np.mean(acclist)) 357 | -------------------------------------------------------------------------------- /qwgc/QWGC_mix.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import random 3 | import copy 4 | 5 | from numpy import pi 6 | from tqdm import trange 7 | from grakel import datasets 8 | from sklearn.model_selection import KFold 9 | 10 | from classifier.qcircuit import ClassifierCircuit 11 | from preprocess.qwfilter import QWfilter 12 | 13 | try: 14 | from utils.notification import Notify 15 | notify = True 16 | except Exception: 17 | notify = False 18 | 19 | THETA_MIN, THETA_MAX = -pi, pi 20 | 21 | 22 | class QWGC: 23 | ''' 24 | In this package, we selected particle swarm optimzier as a optimizer 25 | for tuning vector theta. 26 | You can choose any of optimization way, but gradient way might not 27 | suit for this model. 28 | FIXME: more customizable 29 | In this scheme, the type of coin is constantly changing in each iteraions. 30 | ''' 31 | 32 | def __init__(self, encoder, Cp=1.8, Cg=1.3, n_particle=20, T=100, w=0.8, 33 | ro_max=1.0, n_layer=2, lamb=0.005, n_steps=5, 34 | initial='super', **kwargs): 35 | ''' 36 | Hyper parameters of model. 37 | Input: 38 | FIXME: 39 | encoder: list of binary 40 | e.g. ['01', '10'] 41 | length of encoder must be the same as the number of class 42 | Cp: float (constant value of coefficient 43 | for personal best position) 44 | Cg: float (constant value of coefficient 45 | for grobal best position) 46 | n_particle: int (the number of particles 47 | for searching better parameters) 48 | T: int (the number of iterations of how many times 49 | this model learns) 50 | w: float (constant value of coefficient 51 | for previous particle directions) 52 | ro_max: float (maximum number of random number 53 | which is used for updating parameter) 54 | n_layer: int (the number of layers of mapping circuit) 55 | lamb: float (constant value of coefficient 56 | for the sum of square in error) 57 | n_steps: int (the number of steps of Quantum walk) 58 | ''' 59 | if len(encoder[0]) != n_layer: 60 | raise ValueError('The size of encoder is different\ 61 | from the number of layers') 62 | self.encoder = encoder 63 | self.Cp = Cp 64 | self.Cg = Cg 65 | self.w = w 66 | self.ro_max = ro_max 67 | # FIXME more efficient 68 | if T < 0: 69 | raise ValueError('The number of iterations must be \ 70 | non negative value') 71 | if n_particle < 0: 72 | raise ValueError('The number of particles must be\ 73 | non negative value') 74 | if n_layer <= 0: 75 | raise ValueError('The number of layers must be\ 76 | one or over') 77 | if n_steps < 0: 78 | raise ValueError('The number of steps must be\ 79 | zero or over') 80 | self.T = T 81 | self.n_particle = n_particle 82 | self.layers = n_layer 83 | self.step = n_steps 84 | self.lamb = lamb 85 | self.initial = initial 86 | 87 | def optimize(self, train_data, train_label): 88 | ''' 89 | Input: 90 | train_data: 2dim array (a series of training data) 91 | train_label: 2dim array (a series of label, one hot) 92 | Output: 93 | theta: array 94 | coin_param: ? 95 | ''' 96 | self.n_class = len(train_label[0]) 97 | if self.n_class > 2**self.layers: 98 | raise ValueError('the number of class must be less than 2^layers') 99 | 100 | # initial parameter for Quantum Walk 101 | # each particle has [theta, phi, lambda] 102 | # coin_u3s = np.array([[random.uniform(THETA_MIN, THETA_MAX) 103 | # for i in range(3)] 104 | # for n in range(self.n_particle)]) 105 | coin_u3s = np.array([[random.uniform(THETA_MIN, THETA_MAX), 106 | pi/self.step, pi/self.step] 107 | for n in range(self.n_particle)]) 108 | 109 | ampdata = [QWfilter(coin_u3s[n], self.step, 110 | self.initial).amplitude(train_data) 111 | for n in range(self.n_particle)] 112 | theta_size = self.ceilog(max(set(map(len, ampdata[0])))) 113 | 114 | # FIXME 115 | n_amp = np.array([[self._zero_fill(amp, 2**theta_size) 116 | for amp in ampdata[n]] 117 | for n in range(self.n_particle)]) 118 | 119 | # initialize particles 120 | particles = np.array([[random.uniform(THETA_MIN, THETA_MAX) 121 | for i in range(theta_size)] 122 | for n in range(self.n_particle)]) 123 | velocities = np.array([np.zeros(theta_size) 124 | for n in range(self.n_particle)]) 125 | coin_v = np.array([np.zeros(3) for n in range(self.n_particle)]) 126 | 127 | # for recording best positions of each particle 128 | personal_bpos = copy.copy(particles) 129 | personal_cbpos = copy.copy(coin_u3s) 130 | personal_best_scores = [self._get_cost(amps, train_label, theta) 131 | for amps, theta in zip(n_amp, particles)] 132 | 133 | # recording best position of all particles 134 | best_particle = np.argmin(personal_best_scores) 135 | grobal_best_pos = personal_bpos[best_particle] 136 | grobal_best_coin = coin_u3s[best_particle] 137 | 138 | print('training start!') 139 | # start training 140 | # to check convergence of error, prepare this list 141 | errors = [] 142 | accuracy = [] 143 | for t in trange(self.T, desc='training'): 144 | ampdata = [QWfilter(coin_u3s[n], self.step, 145 | self.initial).amplitude(train_data) 146 | for n in range(self.n_particle)] 147 | n_amp = np.array([[self._zero_fill(amp, 2**theta_size) 148 | for amp in ampdata[n]] 149 | for n in range(self.n_particle)]) 150 | for n in range(self.n_particle): 151 | amp = n_amp[n] 152 | # random number for personal best pos 153 | rnp = random.uniform(0, self.ro_max) 154 | # random number for grobal best 155 | rng = random.uniform(0, self.ro_max) 156 | 157 | # update position 158 | particles[n] = particles[n] + velocities[n] 159 | # update coin pos 160 | 161 | coin_u3s[n] = coin_u3s[n] + coin_v[n] 162 | 163 | velocities[n] = (self.w*velocities[n] + 164 | self.Cp*rnp*(personal_bpos[n]-particles[n]) + 165 | self.Cg*rng*(grobal_best_pos-particles[n])) 166 | 167 | coin_v[n] = (self.w*coin_v[n] + 168 | self.Cp*rnp*(personal_cbpos[n]-coin_u3s[n]) + 169 | self.Cg*rng*(grobal_best_coin-coin_u3s[n])) 170 | 171 | # calculation cost with updated parameters 172 | # and update best position and score 173 | score = self._get_cost(amp, train_label, particles[n]) 174 | if score < personal_best_scores[n]: 175 | personal_best_scores[n] = score 176 | personal_bpos[n] = particles[n] 177 | personal_cbpos[n] = coin_u3s[n] 178 | 179 | # in all particles, calculate which is the best particle 180 | # and coin parameters 181 | best_particle = np.argmin(personal_best_scores) 182 | grobal_best_coin = coin_u3s[best_particle] 183 | grobal_best_pos = personal_bpos[best_particle] 184 | 185 | best_amp_data = QWfilter(grobal_best_coin, self.step, 186 | self.initial).amplitude(train_data) 187 | n_best_amp = [self._zero_fill(amp, 2**theta_size) 188 | for amp in best_amp_data] 189 | error = self._get_cost(n_best_amp, train_label, grobal_best_pos) 190 | accs = self._get_accuracy(n_best_amp, train_label, 191 | grobal_best_pos) 192 | errors.append(error) 193 | accuracy.append(accs) 194 | if t % 10 == 0 and notify: 195 | Notify.notify_error(t, error, accs) 196 | convergence = [errors, accuracy] 197 | return grobal_best_pos, grobal_best_coin, convergence 198 | 199 | def _get_cost(self, data, label, theta): 200 | cost = ClassifierCircuit(data, label, theta, self.n_class, 201 | self.layers, self.encoder).cost() 202 | 203 | error = cost + self.lamb*np.sum([i**2 for i in theta]) 204 | return error 205 | 206 | def _get_accuracy(self, data, label, theta): 207 | answers = ClassifierCircuit(data, label, theta, self.n_class, 208 | self.layers, self.encoder).answers() 209 | acc = self._accs(answers, label) 210 | return acc 211 | 212 | @staticmethod 213 | def ceilog(x): 214 | return int(np.ceil(np.log2(x))) 215 | 216 | def test(self, test_data, theta, coin_param): 217 | ''' 218 | Test function to evaluate the performance 219 | Input: 220 | test_data: 2dim array (a series of test data) 221 | theta: tuned parameters theta 222 | coin_param: optimzed coin parameters 223 | Output: 224 | answers: 2dim array (answers for each data) 225 | ''' 226 | theta_size = len(theta) 227 | ampdata = QWfilter(coin_param, self.step, 228 | self.initial).amplitude(test_data) 229 | n_amp = np.array([self._zero_fill(amp, 2**theta_size) 230 | for amp in ampdata]) 231 | # TODO Inplement the case that the number of class is unknown 232 | answers = ClassifierCircuit(n_amp, None, theta, 2, 233 | self.layers, self.encoder).answers() 234 | return answers 235 | 236 | def _accs(self, ans, label): 237 | count = 0 238 | for i, j in zip(ans, label): 239 | if np.argmax(i) == np.argmax(j): 240 | count += 1 241 | # print('answer ', [np.argmax(i) for i in ans]) 242 | # print('label ', [np.argmax(i) for i in label]) 243 | return count/len(label) 244 | 245 | @staticmethod 246 | def _zero_fill(x, base, array=True): 247 | 248 | # FIXME efficiently 249 | xl = list(x) 250 | x_len = len(xl) 251 | if base - x_len < 0: 252 | raise ValueError('Error') 253 | xs = xl + [0 for _ in range(base-x_len)] 254 | if array: 255 | return np.array(xs) 256 | else: 257 | return xs 258 | 259 | def summary(self, answer, label, printer=False): 260 | accs = self._accs(answer, label) 261 | if printer: 262 | print('The accuracy of this model is ', accs) 263 | return accs 264 | 265 | 266 | def one_hot_encoder(label, n_class): 267 | enc_label = [np.zeros(n_class) for _ in label] 268 | for ilb, lb in enumerate(label): 269 | if lb == -1: 270 | enc_label[ilb][0] = 1 271 | else: 272 | enc_label[ilb][lb] = 1 273 | return enc_label 274 | 275 | 276 | if __name__ == '__main__': 277 | # prepare dataset 278 | import toml 279 | # parsing parameters from toml 280 | config = toml.load('experiments.toml') 281 | p_pso = config['pso'] 282 | p_qw = config['qw'] 283 | 284 | data_name = 'MUTAG' 285 | Data = datasets.fetch_dataset(data_name, verbose=False) 286 | data_x, data_y = np.array(Data.data), np.array(Data.target) 287 | 288 | acclist = [] 289 | k = 5 290 | kf = KFold(n_splits=k, shuffle=True) 291 | 292 | qwgc = QWGC(['01', '10'], Cp=p_pso['Cp'], Cg=p_pso['Cg'], 293 | n_particle=p_pso['particles'], T=p_pso['iterations'], 294 | w=p_pso['w'], ro_max=p_pso['random_max'], 295 | n_layer=p_pso['layers'], lamb=p_pso['lambda'], 296 | n_steps=p_qw['steps'], initial=p_qw['initial']) 297 | for train_index, test_index in kf.split(data_x): 298 | # preprocessing for generating data. 299 | x_train, y_train = data_x[train_index], data_y[train_index] 300 | x_test, y_test = data_x[test_index], data_y[test_index] 301 | 302 | # one hot encoding 303 | y_train = one_hot_encoder(y_train, 2) 304 | y_test = one_hot_encoder(y_test, 2) 305 | 306 | theta, coin_param, conv = qwgc.optimize(x_train, y_train) 307 | # test 308 | ans = qwgc.test(x_test, theta, coin_param) 309 | # evaluate 310 | accs = qwgc.summary(ans, y_test) 311 | 312 | acclist.append(accs) 313 | print(accs) 314 | if notify: 315 | Notify.notify_accs(accs, conv) 316 | print('acclist', acclist) 317 | print('mean', np.mean(acclist)) 318 | -------------------------------------------------------------------------------- /qwgc/QW_kernel.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import random 3 | import matplotlib.pyplot as plt 4 | 5 | from numpy import pi 6 | from tqdm import tqdm, trange 7 | from sklearn.model_selection import KFold 8 | from grakel import datasets 9 | from qiskit import QuantumCircuit, QuantumRegister, ClassicalRegister 10 | from qiskit import Aer, execute 11 | from sklearn import svm 12 | 13 | from preprocess.qwfilter import QWfilter 14 | 15 | try: 16 | from utils.notification import Notify 17 | notify = True 18 | except Exception: 19 | notify = False 20 | 21 | step = 3 22 | THETA_MIN, THETA_MAX = -pi, pi 23 | iteration = 200 24 | backend = Aer.get_backend('qasm_simulator') 25 | shots = 1024 26 | 27 | ''' 28 | This is just very prototype code 29 | ''' 30 | 31 | 32 | def qw_kernel(train_data, train_label, lam=2): 33 | ''' 34 | Input: 35 | train_data: 2dim array (a series of training data) 36 | train_label: 2dim array (a series of label, one hot) 37 | Output: 38 | theta: array 39 | coin_param: ? 40 | ''' 41 | ld = len(train_data) 42 | # start training 43 | # to check convergence of error, prepare this list 44 | weights = np.zeros(ld) 45 | print('training start!') 46 | for i in trange(iteration): 47 | it = random.randint(0, ld-1) 48 | decision = 0 49 | for j in range(it): 50 | # FIXME ambiguous error message 51 | try: 52 | decision += weights[j] * train_label[it] * _kernel_function(train_data[it], train_data[j], 7) 53 | except ValueError: 54 | continue 55 | decision *= train_label[it]/lam 56 | if decision < 1: 57 | weights[it] += 1 58 | return weights 59 | 60 | 61 | def test(x_train, y_train, x_test, y_test, weights): 62 | print('test start!') 63 | errors = 0 64 | for ila, lb_test in tqdm(enumerate(y_test)): 65 | decision = 0 66 | for ilb, lb_train in enumerate(y_train): 67 | decision += weights[ilb]*y_train[ilb]*_kernel_function(x_train[ilb], x_test[ila], 7) 68 | if decision < 0: 69 | prediction = -1 70 | else: 71 | prediction = 1 72 | if prediction != y_test[ila]: 73 | errors += 1 74 | return 1 - errors/len(y_test) 75 | 76 | 77 | def _kernel_function(x, y, qsize): 78 | # definition of coin 79 | coin_u3s = np.array([pi, 0, pi/2]) 80 | 81 | ampdata_x = QWfilter(coin_u3s, step, 'super').single_amplitude(x) 82 | x_amp = _zero_fill(ampdata_x, 2**qsize) 83 | 84 | q1 = QuantumRegister(qsize) 85 | qc1 = QuantumCircuit(q1, name='QW1') 86 | qc1.initialize(x_amp, q1) 87 | qw1 = qc1.to_instruction() 88 | 89 | ampdata_y = QWfilter(coin_u3s, step, 'super').single_amplitude(y) 90 | y_amp = _zero_fill(ampdata_y, 2**qsize) 91 | 92 | q2 = QuantumRegister(qsize) 93 | qc2 = QuantumCircuit(q2) 94 | qc2.initialize(y_amp, q2) 95 | qw2 = qc2.to_instruction() 96 | 97 | kq = QuantumRegister(qsize) 98 | c = ClassicalRegister(qsize) 99 | kqc = QuantumCircuit(kq, c) 100 | kqc.append(qw1, qargs=kq) 101 | kqc.append(qw2, qargs=kq) 102 | kqc.measure(kq, c) 103 | # calc prob '000...0' 104 | job = execute(kqc, backend=backend, shots=shots) 105 | count = job.result().get_counts(kqc) 106 | return count.get('0'*qsize, 0)/shots 107 | 108 | 109 | def ceilog(x): 110 | return int(np.ceil(np.log2(x))) 111 | 112 | 113 | def _zero_fill(x, base, array=True): 114 | # FIXME efficiently 115 | xl = list(x) 116 | x_len = len(xl) 117 | if base - x_len < 0: 118 | raise ValueError('Error') 119 | xs = xl + [0 for _ in range(base-x_len)] 120 | if array: 121 | return np.array(xs) 122 | else: 123 | return xs 124 | 125 | 126 | if __name__ == '__main__': 127 | data_name = 'MUTAG' 128 | Data = datasets.fetch_dataset(data_name, verbose=False) 129 | data_x, data_y = np.array(Data.data), np.array(Data.target) 130 | 131 | k = 5 132 | kf = KFold(n_splits=k, shuffle=True) 133 | accuracy = [] 134 | for train_index, test_index in kf.split(data_x): 135 | # preprocessing for generating data. 136 | x_train, y_train = data_x[train_index], data_y[train_index] 137 | x_test, y_test = data_x[test_index], data_y[test_index] 138 | weight = qw_kernel(x_train, y_train) 139 | accs = test(x_train, y_train, x_test, y_test, weight) 140 | print(accs) 141 | if notify: 142 | Notify.notify_accs(accs, 'svm') 143 | accuracy.append(accs) 144 | Notify.notify_accs(accuracy, 'K5 result') 145 | Notify.notify_accs(np.mean(accuracy), 'K5 result mean') 146 | print(accuracy) 147 | print(np.mean(accuracy)) 148 | -------------------------------------------------------------------------------- /qwgc/__init__.py: -------------------------------------------------------------------------------- 1 | from .QWGC import QWGC 2 | -------------------------------------------------------------------------------- /qwgc/classifier/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qwqmlf/qwgc/d06c805023bf37a1252505f7dc6e30461d861440/qwgc/classifier/__init__.py -------------------------------------------------------------------------------- /qwgc/classifier/costfunc.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qwqmlf/qwgc/d06c805023bf37a1252505f7dc6e30461d861440/qwgc/classifier/costfunc.py -------------------------------------------------------------------------------- /qwgc/classifier/qcircuit.py: -------------------------------------------------------------------------------- 1 | from qiskit import QuantumCircuit, QuantumRegister, ClassicalRegister 2 | from qiskit import Aer, execute 3 | from scipy.special import softmax 4 | 5 | from utils.notification import Notify 6 | import numpy as np 7 | from numpy import pi 8 | 9 | QASM = Aer.get_backend('qasm_simulator') 10 | 11 | 12 | class ClassifierCircuit: 13 | ''' 14 | This class is the core part of this algorithm. 15 | FIXME inherit QWGC and take super constructor. 16 | ''' 17 | 18 | def __init__(self, data, label, theta, n_class, 19 | layers, encoder, shots=1024): 20 | ''' 21 | Input: 22 | encoding: list (discribe the correspondence of 23 | measurement basis and label) 24 | e.g. ['01', '10'] 25 | ''' 26 | self.data = data 27 | if label is None: 28 | # FIXME this is for test 29 | self.label = 0 30 | else: 31 | if len(label[0]) != len(encoder): 32 | raise ValueError('dimension of label and \ 33 | encoder must be the same') 34 | self.label = label 35 | # FIXME 36 | self.n_class = n_class 37 | self.theta = theta 38 | self.qsize = len(theta) 39 | self.layers = layers 40 | self.encoder = encoder 41 | self.shots = shots 42 | if layers > self.qsize: 43 | raise Warning('The length of theta is shorter \ 44 | than the number of layers') 45 | 46 | def _circuit_constructor(self, visualize=False, notify=False): 47 | ''' 48 | returning circuits 49 | TODO 50 | if visualize got the list of integers, 51 | then, print out the circuit of certain index 52 | ''' 53 | qcs = [] 54 | nq = self.qsize 55 | layer = self.layers 56 | if notify: 57 | Notify.notify_accs(list(self.data[0]), list(self.data[-1])) 58 | for index, d in enumerate(self.data): 59 | # qubits for representing data 60 | qr = QuantumRegister(nq) 61 | # qubits for mapping data 62 | mp = QuantumRegister(layer) 63 | c = ClassicalRegister(layer) 64 | qc = QuantumCircuit(qr, mp, c, name='data%d' % index) 65 | qc.initialize(d, qr) 66 | # qc.h(qr) 67 | qc = self._map(qc, qr, mp) 68 | qc.measure(mp, c) 69 | qcs.append(qc) 70 | return qcs 71 | 72 | def _map(self, qc, qr, mp): 73 | # counter = 0 74 | for ith, theta in enumerate(self.theta): 75 | # if ith % self.layers == 0: 76 | qc.cry(theta, qr[ith], mp[ith % self.layers]) 77 | # else: 78 | # qc.h(mp[ith%self.layers]) 79 | # qc.cu3(theta, pi/2, pi/2, qr[ith], mp[ith % self.layers]) 80 | # qc.h(mp[ith%self.layers]) 81 | return qc 82 | 83 | def cost(self, notify=False): 84 | ''' 85 | This function is the interface to pass through 86 | the result of measurement 87 | Input: 88 | data: amplitude vector 89 | Output: 90 | result of measurement 91 | ''' 92 | qcs = self._circuit_constructor(notify=notify) 93 | probs = self._get_result(qcs, notify=notify) 94 | cross = np.mean([self._cross_entropy_error(pb, lb) 95 | for pb, lb in zip(probs, self.label)]) 96 | return cross 97 | 98 | def answers(self): 99 | qcs = self._circuit_constructor() 100 | probs = self._get_result(qcs) 101 | answers = [np.zeros(self.n_class) for _ in probs] 102 | for ipb, pb in enumerate(probs): 103 | ind = np.argmax(pb) 104 | answers[ipb][ind] = 1 105 | return answers 106 | 107 | def _get_result(self, qcs, notify=False): 108 | ''' 109 | returning probabilities of estimation 110 | ''' 111 | job = execute(qcs, backend=QASM, shots=self.shots) 112 | counts = [job.result().get_counts(qc) for qc in qcs] 113 | dinom = [sum([cs.get(i, 0) for i in self.encoder]) for cs in counts] 114 | if notify: 115 | bins = [format(i, "02b") for i in range(4)] 116 | Notify.notify_accs("After Classify data0", [counts[0].get(b, 0)/sum(counts[0].values()) for b in bins]) 117 | Notify.notify_accs("After Classify data last", [counts[-1].get(b, 0)/sum(counts[-1].values()) for b in bins]) 118 | enc_probs = [np.array([cs.get(i, 0)/(din+1e-10) for i in self.encoder]) 119 | for cs, din in zip(counts, dinom)] 120 | return enc_probs 121 | 122 | @staticmethod 123 | def ceilog(x): 124 | return int(np.ceil(np.log2(x))) 125 | 126 | @staticmethod 127 | def _cross_entropy_error(y, t, delta=1e-7): 128 | return -np.sum(t * np.log(y + delta)) 129 | -------------------------------------------------------------------------------- /qwgc/experiments.toml: -------------------------------------------------------------------------------- 1 | title = 'Hypter parameters of qwgc' 2 | 3 | [pso] 4 | Cp = 1.4 5 | Cg = 1.1 6 | particles = 10 7 | iterations = 200 8 | w = 0.7 9 | random_max = 1.2 10 | layers = 6 11 | lambda = 0.0005 12 | 13 | [qw] 14 | steps = 5 15 | initial = 'super' 16 | -------------------------------------------------------------------------------- /qwgc/preprocess/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/qwqmlf/qwgc/d06c805023bf37a1252505f7dc6e30461d861440/qwgc/preprocess/__init__.py -------------------------------------------------------------------------------- /qwgc/preprocess/gparse.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | 3 | 4 | class GraphInfo: 5 | def __init__(self, adjacency_matrix): 6 | self.adja = adjacency_matrix 7 | self.dim = len(adjacency_matrix) 8 | 9 | def shift(self): 10 | deg = self._degrees 11 | node_index = self._first_node() 12 | ndim = self.dim 13 | size = node_index[ndim-1] + len(deg[ndim-1]) 14 | deg2 = [[0] for _ in deg] 15 | for i, v in enumerate(deg): 16 | deg2[i] = list(set(v)) 17 | array = np.zeros((size, size), dtype=int) 18 | for i, val in enumerate(deg): 19 | index1 = node_index[i] 20 | n = 0 21 | for j, jval in enumerate(deg2[i]): 22 | nolinks = self.adja[jval][i] 23 | n = self._coinst(i, val[j]) 24 | coinst = index1 + n 25 | node = val[j] 26 | for k in deg2[node]: 27 | if k == i: 28 | coinst2 = node_index[node] + self._coinst(node, i) 29 | for k in range(int(nolinks)): 30 | array[coinst+k][coinst2+k] = 1 31 | # FIXME catch before 32 | for i, v in enumerate(array): 33 | for j, w in enumerate(array): 34 | if array[i][j] == 1: 35 | array[j][i] = 1 36 | return array 37 | 38 | @property 39 | def _degrees(self): 40 | degrees = [[] for _ in range(self.dim)] 41 | for iv, v in enumerate(self.adja): 42 | for ix, x in enumerate(v): 43 | if x == 1: 44 | degrees[iv].append(ix) 45 | return degrees 46 | 47 | def _first_node(self): 48 | n = 0 49 | array = np.zeros(len(self._degrees), dtype=int) 50 | for i, v in enumerate(self._degrees): 51 | array[i] = n 52 | n += len(v) 53 | return array 54 | 55 | def _coinst(self, i, j): 56 | n1 = 0 57 | n2 = 0 58 | idim = len(self._degrees[i]) 59 | for k in range(idim): 60 | if n2 == 0: 61 | if self._degrees[i][k] != j: 62 | n1 += 1 63 | else: 64 | n2 = 1 65 | if n1 == idim: 66 | return 'no link' 67 | else: 68 | return n1 69 | -------------------------------------------------------------------------------- /qwgc/preprocess/qwalk.py: -------------------------------------------------------------------------------- 1 | ''' 2 | Thank you Barr Katie for providing your code. 3 | I reffered this paper and code. 4 | Barr, Katie, Toby Fleming, and Viv Kendon. 5 | "Simulation methods for quantum walks on 6 | graphs applied to perfect state transfer and 7 | formal language recognition." CoSMoS 2013 (2013): 1. 8 | ''' 9 | import numpy as np 10 | import numpy.linalg as la 11 | from .gparse import GraphInfo 12 | 13 | 14 | class QuantumWalk: 15 | 16 | def __init__(self, initial_state, coin_operator, adjacency_matrix, 17 | tolerance=1e-5): 18 | ''' 19 | initial_state: None or array like 20 | coin_operator: array like 21 | adjacency_matrix: array like 22 | ''' 23 | gf = GraphInfo(adjacency_matrix) 24 | if not is_unitary(coin_operator): 25 | raise ValueError('Coin operator must be unitary') 26 | else: 27 | self.coin_operator = coin_operator 28 | self.shift_operator = gf.shift() 29 | self.time_ev = np.dot(self.shift_operator, coin_operator) 30 | eigs, _ = la.eig(self.time_ev) 31 | if initial_state is None: 32 | self.create_default_initial() 33 | else: 34 | self.initial_state = initial_state 35 | self.current_state = initial_state 36 | self.adjacency = adjacency_matrix 37 | self.dim = len(adjacency_matrix) 38 | self.tolerance = tolerance 39 | 40 | def create_default_initial(self): 41 | basis_state = self.time_ev.shape[0] 42 | vec = np.zeros((basis_state)) 43 | vec[0] = 1 44 | self.initial_state = vec 45 | 46 | def step(self): 47 | self.current_state = np.dot(self.time_ev, self.current_state) 48 | 49 | def step_back(self): 50 | self.current_state = np.dot(np.concatenate(self.time_ev.transpose()), 51 | self.current_state) 52 | 53 | def steps(self, n): 54 | for i in range(n): 55 | state = np.dot(self.time_ev, self.current_state) 56 | self.current_state = state 57 | 58 | @property 59 | def node_deg(self): 60 | return [int(np.sum(self.adjacency[i])) for i in range(self.dim)] 61 | 62 | def prob_at_node(self, index): 63 | if index > self.dim: 64 | raise ValueError('Graph does not have %d nodes' % index) 65 | probs = self.calc_probs 66 | return probs[index] 67 | 68 | def calc_probs(self): 69 | probs = np.zeros(self.dim) 70 | ind = 0 71 | for i in range(self.dim): 72 | for j in range(self.node_deg[i]): 73 | amps_at_j = self.current_state[ind] 74 | probs[i] += amps_at_j * np.conjugate(amps_at_j) 75 | ind += 1 76 | assert np.isclose(np.sum(probs), 1, atol=self.tolerance) 77 | return probs 78 | 79 | def calc_amp(self): 80 | # FIXME in this implementation, we can't get amplitudes of each node 81 | return self.current_state 82 | 83 | def n_steps(self, steps): 84 | if steps < 0: 85 | raise ValueError('steps must be 0 or over') 86 | elif steps == 0: 87 | self.current_state = self.initial_state 88 | return self.current_state 89 | 90 | eig = la.eig(self.time_ev)[1] 91 | inverse = la.inv(eig) 92 | diag = np.dot(np.dot(inverse, self.time_ev), eig) 93 | 94 | for i, _ in enumerate(diag): 95 | transition = diag[i][i] 96 | x = transition.real 97 | y = transition.imag 98 | theta = np.arctan2(y, x) 99 | diag[i][i] = np.cos(steps*theta) + complex(0, np.sin(steps*theta)) 100 | transform_bvec = np.dot(inverse, self.initial_state) 101 | evolved = np.dot(diag, transform_bvec) 102 | trans_back = np.dot(eig, evolved) 103 | self.current_state = trans_back 104 | return trans_back 105 | 106 | 107 | def is_unitary(operator): 108 | h, w = operator.shape 109 | if not h == w: 110 | return False 111 | adjoint = np.conjugate(operator.transpose()) 112 | product1 = np.dot(operator, adjoint) 113 | product2 = np.dot(adjoint, operator) 114 | ida = np.eye(h) 115 | return np.allclose(product1, ida) & np.allclose(product2, ida) 116 | -------------------------------------------------------------------------------- /qwgc/preprocess/qwfilter.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import random 3 | import copy 4 | import itertools 5 | 6 | from grakel import Graph 7 | from .qwalk import QuantumWalk 8 | 9 | np.set_printoptions(linewidth=100000) 10 | 11 | 12 | class QWfilter: 13 | def __init__(self, u3param, step, initial): 14 | self.u3p = u3param 15 | self.step = step 16 | self.initial = initial 17 | 18 | def amplitude(self, data, normalize=False): 19 | ''' 20 | interface 21 | Output: 22 | amplitude (no normalize) 23 | ''' 24 | adjacency = [Graph(d[0]).get_adjacency_matrix() for d in data] 25 | amplitude = [] 26 | for ad in adjacency: 27 | amp = self.coin_walk(ad) 28 | # amp = self.szegedy_google(ad) 29 | amplitude.append(amp) 30 | return amplitude 31 | 32 | def coin_walk(self, ad): 33 | count = np.count_nonzero(ad)//2 34 | nad = len(ad) 35 | # prepare coin with u3 parameter 36 | coin = self._compose_coins(count, ad) 37 | # prepare initial state 38 | initial_state = self._initial(count, nad) 39 | 40 | # construct quantum walk and go n steps 41 | qwalk = QuantumWalk(initial_state, coin, ad) 42 | qwalk.n_steps(self.step) 43 | amplitude = qwalk.calc_amp() 44 | # print(list(amplitude)) 45 | return amplitude 46 | 47 | def szegedy_google(self, adjacency): 48 | """ 49 | quantum pagerank 50 | """ 51 | G = google_matrix(adjacency) 52 | initial = 1/2*np.array([np.sqrt(G[j][i]) 53 | for i, _ in enumerate(G) 54 | for j, _ in enumerate(G)]) 55 | print(sum([abs(i)**2 for i in initial])) 56 | Pi_op = self._Pi_operator(G) 57 | print(is_unitary(Pi_op)) 58 | swap = self._swap_operator(len(G)) 59 | print(is_unitary(swap)) 60 | operator = (2*Pi_op) - np.identity(len(Pi_op)) 61 | Szegedy = np.dot(operator, swap) 62 | Szegedy_n = copy.deepcopy(Szegedy) 63 | if self.step == 0: 64 | return initial 65 | elif self.step == 1: 66 | amp = np.dot(Szegedy, self.initial) 67 | return amp 68 | else: 69 | for n in range(self.step-1): 70 | Szegedy_n = np.dot(Szegedy_n, Szegedy) 71 | amp = np.dot(Szegedy_n, initial) 72 | return amp 73 | 74 | def _Pi_operator(self, ptran): 75 | ''' 76 | This is not a quantum operation, 77 | just returning matrix 78 | ''' 79 | lg = len(ptran) 80 | psi_op = [] 81 | count = 0 82 | for i in range(lg): 83 | psi_vec = [0 for _ in range(lg**2)] 84 | for j in range(lg): 85 | psi_vec[count] = np.sqrt(ptran[j][i]) 86 | count += 1 87 | psi_op.append(np.kron(np.array(psi_vec).T, 88 | np.conjugate(psi_vec)).reshape((lg**2, lg**2))) 89 | Pi = psi_op[0] 90 | for i in psi_op[1:]: 91 | Pi = np.add(Pi, i) 92 | return Pi 93 | 94 | def _swap_operator(self, lad): 95 | # find closest 2 pow 96 | base = int(np.ceil(np.log2((1 << int(np.ceil(np.log2(lad))))))) 97 | swap = np.zeros((lad**2, lad**2)) 98 | for i in range(lad): 99 | # ibin = format(i, '0%db' % base) 100 | for j in range(lad): 101 | # jbin = format(j, '0%db' % base) 102 | # a = int(ibin + jbin, 2) 103 | # b = int(jbin + ibin, 2) 104 | ai = np.array([1 if t == i else 0 for t in range(lad)]) 105 | bi = np.conjugate(np.array([1 if k == j else 0 for k in range(lad)]).T) 106 | swap += np.kron(ai, bi) 107 | # raise Exception("") 108 | return swap 109 | 110 | def _reform(self, ad): 111 | for il, ln in enumerate(ad): 112 | rd = random.choice([i for i, _ in enumerate(ln) if i != il]) 113 | if sum(ln) == 1: 114 | ad[il][rd] = 1 115 | return ad 116 | 117 | def single_amplitude(self, d): 118 | ad = Graph(d[0]).get_adjacency_matrix() 119 | count = np.count_nonzero(ad)//2 120 | nad = len(ad) 121 | 122 | # prepare coin with u3 parameter 123 | coin = self._compose_coins(count, ad) 124 | # prepare initial state 125 | initial_state = self._initial(count, nad) 126 | 127 | # construct quantum walk and go n steps 128 | qwalk = QuantumWalk(initial_state, coin, ad) 129 | qwalk.n_steps(self.step) 130 | amplitude = qwalk.calc_amp() 131 | return amplitude 132 | 133 | def single_prob(self, d): 134 | ad = Graph(d[0]).get_adjacency_matrix() 135 | count = np.count_nonzero(ad)//2 136 | nad = len(ad) 137 | 138 | # prepare coin with u3 parameter 139 | coin = self._compose_coins(count, ad) 140 | # prepare initial state 141 | initial_state = self._initial(count, nad) 142 | 143 | # construct quantum walk and go n steps 144 | qwalk = QuantumWalk(initial_state, coin, ad) 145 | qwalk.n_steps(self.step) 146 | probability = qwalk.calc_probs() 147 | return probability 148 | 149 | def _compose_coins(self, count, adja): 150 | ''' 151 | Input: 152 | count: the number of non-zero elements in adjacency matrix 153 | adja: adjacency matrix of data 154 | pahse: FIXME default True 155 | useing quantum coin with phase 156 | Output: 157 | coin: 2d matrix (unitary) 158 | ''' 159 | co = [] 160 | elcoin = [] 161 | coin = np.array(np.diag(np.zeros(count*2)), dtype=np.complex) 162 | for lad in adja: 163 | s = int(sum(lad)) 164 | co.append(s) 165 | section = [0] + list(itertools.accumulate(co))[0:-1] 166 | for c in co: 167 | coin_element = self._coin(c) 168 | for ce in coin_element: 169 | elcoin.append(ce) 170 | counter = 0 171 | for ic, coinel in enumerate(zip(coin, elcoin)): 172 | ci, ce = coinel[0], coinel[1] 173 | nce = len(ce) 174 | ci[counter:counter+nce] = ce 175 | if ic+1 in section: 176 | counter = ic+1 177 | if not is_unitary(coin): 178 | raise Exception('coin operator must be unitary') 179 | return coin 180 | 181 | def _coin(self, num): 182 | if num == 2: 183 | # testing if this is good enough to make good amp or not 184 | coin = self.U3(self.u3p[0], self.u3p[1], self.u3p[2]) 185 | else: 186 | coin = np.array([[2/num for k in range(num)] 187 | for i in range(num)] - np.identity(num)) 188 | 189 | if not is_unitary(coin): 190 | raise Exception("elementary operator must be unitary.") 191 | return coin 192 | 193 | def _initial(self, count, nad): 194 | if self.initial is None: 195 | initial_state = None 196 | elif isinstance(self.initial, list or np.ndarray): 197 | initial_state = self.initial 198 | else: 199 | initial_state = [1/np.sqrt(nad) for i in range(nad)] + \ 200 | [0 for i in range(2*count-nad)] 201 | # FIXME check threshold 202 | assert(np.sum(i**2 for i in initial_state)-1 < 1e-5) 203 | return initial_state 204 | 205 | @staticmethod 206 | def U3(theta, phi, lamb): 207 | return np.array([[np.cos(theta/2), 208 | -np.exp(1j*lamb)*np.sin(theta/2)], 209 | [np.exp(1j*phi)*np.sin(theta/2), 210 | np.exp(1j*lamb+1j*phi)*np.cos(theta/2)]]) 211 | 212 | 213 | def is_unitary(operator, tolerance=0.0001): 214 | h, w = operator.shape 215 | if not h == w: 216 | return False 217 | adjoint = np.conjugate(operator.transpose()) 218 | product1 = np.dot(operator, adjoint) 219 | product2 = np.dot(adjoint, operator) 220 | ida = np.eye(h) 221 | return np.allclose(product1, ida) & np.allclose(product2, ida) 222 | 223 | 224 | def prob_transition(graph, gtype='google'): 225 | if gtype == 'google': 226 | return google_matrix(graph) 227 | else: 228 | pmatrix = np.zeros(graph.shape) 229 | indegrees = np.sum(graph, axis=0) 230 | for ix, indeg in enumerate(indegrees): 231 | if indeg == 0: 232 | pmatrix[:, ix] = graph[:, ix] 233 | else: 234 | pmatrix[:, ix] = graph[:, ix]/indeg 235 | return pmatrix 236 | 237 | 238 | def google_matrix(graph, alpha=0.85): 239 | E = np.zeros((len(graph), len(graph))) 240 | for i, t in enumerate(graph): 241 | if sum(graph[:, i]) == 0: 242 | E[:, i] = np.array([1/len(graph) for _ in graph]) 243 | else: 244 | for ij, j in enumerate(t): 245 | E[ij, i] = j/sum(graph[:, i]) 246 | G = alpha*E + (1-alpha)/(len(graph)) * np.ones((len(graph), len(graph))) 247 | return G 248 | -------------------------------------------------------------------------------- /qwgc/utils/.gitignore: -------------------------------------------------------------------------------- 1 | notification.py -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | # 2 | # These requirements were autogenerated by pipenv 3 | # To regenerate from the project's Pipfile, run: 4 | # 5 | # pipenv lock --requirements 6 | # 7 | 8 | -i https://pypi.org/simple/ 9 | attrs==21.2.0; python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4' 10 | autopep8==1.5.4 11 | cached-property==1.5.2; python_version < '3.8' 12 | certifi==2021.5.30 13 | cffi==1.14.6 14 | charset-normalizer==2.0.4; python_version >= '3' 15 | cryptography==3.3.2 16 | cycler==0.10.0 17 | cython==0.29.24; python_version >= '2.6' and python_version not in '3.0, 3.1, 3.2, 3.3' 18 | dill==0.3.4; python_version >= '2.7' and python_version != '3.0' 19 | dlx==1.0.4 20 | docplex==2.21.207 21 | fastdtw==0.3.4 22 | fastjsonschema==2.15.1 23 | future==0.18.2; python_version >= '2.6' and python_version not in '3.0, 3.1, 3.2, 3.3' 24 | grakel-dev==0.1a6 25 | grakel==0.1b7 26 | h5py==3.4.0; python_version >= '3.7' 27 | idna==3.2; python_version >= '3' 28 | importlib-metadata==4.8.1; python_version < '3.8' 29 | inflection==0.5.1; python_version >= '3.5' 30 | joblib==1.0.1; python_version >= '3.6' 31 | jsonschema==3.2.0 32 | kiwisolver==1.3.2; python_version >= '3.7' 33 | llvmlite==0.37.0; python_version < '3.10' and python_version >= '3.7' 34 | lxml==4.6.3; python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4' 35 | matplotlib==3.4.3; python_version >= '3.7' 36 | more-itertools==8.9.0; python_version >= '3.5' 37 | mpmath==1.2.1 38 | multitasking==0.0.9 39 | nest-asyncio==1.5.1; python_version >= '3.5' 40 | networkx==2.6.2; python_version >= '3.7' 41 | nose==1.3.7 42 | ntlm-auth==1.5.0; python_version >= '2.6' and python_version not in '3.0, 3.1, 3.2, 3.3' 43 | numba==0.54.0; python_version < '3.10' and python_version >= '3.7' 44 | numpy==1.20.3; python_version == '3.7' 45 | pandas==1.3.2; python_full_version >= '3.7.1' 46 | pillow==8.3.2; python_version >= '3.6' 47 | ply==3.11 48 | psutil==5.8.0; python_version >= '2.6' and python_version not in '3.0, 3.1, 3.2, 3.3' 49 | pybind11==2.7.1; python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4' 50 | pycodestyle==2.7.0; python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3' 51 | pycparser==2.20; python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3' 52 | pyparsing==2.4.7; python_version >= '2.6' and python_version not in '3.0, 3.1, 3.2, 3.3' 53 | pyrsistent==0.18.0; python_version >= '3.6' 54 | python-constraint==1.4.0 55 | python-dateutil==2.8.2; python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3' 56 | pytz==2021.1 57 | qiskit-aer==0.7.0; python_version >= '3.6' 58 | qiskit-aqua==0.8.0; python_version >= '3.6' 59 | qiskit-ibmq-provider==0.11.0; python_version >= '3.6' 60 | qiskit-ignis==0.5.0; python_version >= '3.6' 61 | qiskit-terra==0.16.0; python_version >= '3.6' 62 | qiskit==0.23.0 63 | quandl==3.6.1; python_version >= '3.5' 64 | requests-ntlm==1.1.0 65 | requests==2.26.0; python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4, 3.5' 66 | retworkx==0.10.1; python_version >= '3.6' 67 | scikit-learn==0.24.2; python_version >= '3.6' 68 | scipy==1.7.1; python_version < '3.10' and python_version >= '3.7' 69 | seaborn==0.11.0 70 | six==1.16.0; python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3' 71 | slackweb==1.0.5 72 | sympy==1.8; python_version >= '3.6' 73 | threadpoolctl==2.2.0; python_version >= '3.6' 74 | toml==0.10.2 75 | tqdm==4.51.0 76 | typing-extensions==3.10.0.2; python_version < '3.8' 77 | umap-learn==0.4.6 78 | urllib3==1.26.6; python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4' and python_version < '4' 79 | websockets==9.1; python_full_version >= '3.6.1' 80 | yfinance==0.1.63 81 | zipp==3.5.0; python_version >= '3.6' 82 | -------------------------------------------------------------------------------- /setup.py: -------------------------------------------------------------------------------- 1 | from setuptools import setup 2 | 3 | with open('requirements.txt', 'r') as f: 4 | requirements = [line.strip() for line in f] 5 | 6 | discription = "qwgc is a quantum walk graph classifier \ 7 | for classification for Graph data." 8 | setup( 9 | name="qwgc", 10 | version="0.0.3", 11 | description="Graph classifier based on quantum walk", 12 | long_description=discription, 13 | url="https://Chibikuri.github.io/qwgc", 14 | author="Ryosuke Satoh", 15 | author_email="ryosuke.satoh.wk@gmail.com", 16 | license="Apache 2.0", 17 | classifiers=[ 18 | "License :: OSI Approved :: Apache Software License", 19 | "Operating System :: MacOS", 20 | "Operating System :: POSIX :: Linux", 21 | "Programming Language :: Python :: 3.6", 22 | "Programming Language :: Python :: 3.7", 23 | ], 24 | keywords="quantum walk machine learning", 25 | install_requires=requirements, 26 | include_package_data=True, 27 | python_requires=">=3.5", 28 | ) 29 | -------------------------------------------------------------------------------- /test/test_QWGC.py: -------------------------------------------------------------------------------- 1 | import pytest 2 | --------------------------------------------------------------------------------