├── .github └── workflows │ └── python-publish.yml ├── .gitignore ├── LICENSE ├── README.md ├── deep_linear_network ├── __init__.py └── deep_linear_network.py ├── diagram.png └── setup.py /.github/workflows/python-publish.yml: -------------------------------------------------------------------------------- 1 | # This workflows will upload a Python Package using Twine when a release is created 2 | # For more information see: https://help.github.com/en/actions/language-and-framework-guides/using-python-with-github-actions#publishing-to-package-registries 3 | 4 | name: Upload Python Package 5 | 6 | on: 7 | release: 8 | types: [created] 9 | 10 | jobs: 11 | deploy: 12 | 13 | runs-on: ubuntu-latest 14 | 15 | steps: 16 | - uses: actions/checkout@v2 17 | - name: Set up Python 18 | uses: actions/setup-python@v2 19 | with: 20 | python-version: '3.x' 21 | - name: Install dependencies 22 | run: | 23 | python -m pip install --upgrade pip 24 | pip install setuptools wheel twine 25 | - name: Build and publish 26 | env: 27 | TWINE_USERNAME: ${{ secrets.PYPI_USERNAME }} 28 | TWINE_PASSWORD: ${{ secrets.PYPI_PASSWORD }} 29 | run: | 30 | python setup.py sdist bdist_wheel 31 | twine upload dist/* 32 | -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | # Byte-compiled / optimized / DLL files 2 | __pycache__/ 3 | *.py[cod] 4 | *$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 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | MIT License 2 | 3 | Copyright (c) 2020 Phil Wang 4 | 5 | Permission is hereby granted, free of charge, to any person obtaining a copy 6 | of this software and associated documentation files (the "Software"), to deal 7 | in the Software without restriction, including without limitation the rights 8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 9 | copies of the Software, and to permit persons to whom the Software is 10 | furnished to do so, subject to the following conditions: 11 | 12 | The above copyright notice and this permission notice shall be included in all 13 | copies or substantial portions of the Software. 14 | 15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 21 | SOFTWARE. 22 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | 2 | 3 | ## Deep Linear Network - Pytorch 4 | 5 | A simple to use deep linear network module. Useful for matrix factorization or for passing an input tensor through a series of square weight matrices, where it was discovered that gradient descent implicitly regularizes the output to low-rank solutions. 6 | 7 | LeCun's paper uses this unique property to optimize the latent of an autoencoder to be low-rank. 8 | 9 | The module will take care of collapsing the linear weight matrices into one weight matrix, caching it across evaluation calls (but expired on training). 10 | 11 | ## Install 12 | 13 | ```bash 14 | $ pip install deep-linear-network 15 | ``` 16 | 17 | ## Usage 18 | 19 | Matrix factorization 20 | 21 | ```python 22 | import torch 23 | from deep_linear_network import DeepLinear 24 | 25 | x = torch.randn(1, 1024, 256) 26 | linear = DeepLinear(256, 10, 512) # w1 (256 x 10) @ w2 (10 x 512) 27 | linear(x) # (1, 1024, 512) 28 | ``` 29 | 30 | Deep Linear Network 31 | 32 | ```python 33 | import torch 34 | from deep_linear_network import DeepLinear 35 | 36 | x = torch.randn(1, 1024, 256) 37 | linear = DeepLinear(256, 256, 256, 256, 128) # w1-w3 (256 x 256) w4 (256 x 128) 38 | linear(x) # (1, 1024, 128) 39 | ``` 40 | 41 | ## Citations 42 | 43 | ```bibtex 44 | @misc{arora2019implicit, 45 | title={Implicit Regularization in Deep Matrix Factorization}, 46 | author={Sanjeev Arora and Nadav Cohen and Wei Hu and Yuping Luo}, 47 | year={2019}, 48 | eprint={1905.13655}, 49 | archivePrefix={arXiv}, 50 | primaryClass={cs.LG} 51 | } 52 | ``` 53 | 54 | ```bibtex 55 | @misc{jing2020implicit, 56 | title={Implicit Rank-Minimizing Autoencoder}, 57 | author={Li Jing and Jure Zbontar and Yann LeCun}, 58 | year={2020}, 59 | eprint={2010.00679}, 60 | archivePrefix={arXiv}, 61 | primaryClass={cs.LG} 62 | } 63 | ``` 64 | -------------------------------------------------------------------------------- /deep_linear_network/__init__.py: -------------------------------------------------------------------------------- 1 | from deep_linear_network.deep_linear_network import DeepLinear 2 | -------------------------------------------------------------------------------- /deep_linear_network/deep_linear_network.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from torch import nn 3 | from functools import reduce 4 | 5 | def mm(x, y): 6 | return x @ y 7 | 8 | class DeepLinear(nn.Module): 9 | def __init__(self, dim_in, *dims): 10 | super().__init__() 11 | dims = [dim_in, *dims] 12 | pairs = list(zip(dims[:-1], dims[1:])) 13 | weights = list(map(lambda d: nn.Parameter(torch.randn(d)), pairs)) 14 | self.weights = nn.ParameterList(weights) 15 | self._cache = None 16 | 17 | def forward(self, x): 18 | if self.training: 19 | self._cache = None 20 | return reduce(mm, self.weights, x) 21 | 22 | if self._cache is not None: 23 | return x @ self._cache 24 | 25 | head, *tail = self.weights 26 | weight = reduce(mm, tail, head) 27 | self._cache = weight 28 | return x @ weight 29 | -------------------------------------------------------------------------------- /diagram.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/lucidrains/deep-linear-network/b2fa7f1ec21159a88cfeacde1c7e5f7c063764a9/diagram.png -------------------------------------------------------------------------------- /setup.py: -------------------------------------------------------------------------------- 1 | from setuptools import setup, find_packages 2 | 3 | setup( 4 | name = 'deep-linear-network', 5 | packages = find_packages(), 6 | version = '0.0.1', 7 | license='MIT', 8 | description = 'Deep Linear Network - Pytorch', 9 | author = 'Phil Wang', 10 | author_email = 'lucidrains@gmail.com', 11 | url = 'https://github.com/lucidrains/deep-linear-network', 12 | keywords = [ 13 | 'artificial intelligence', 14 | 'attention mechanism', 15 | ], 16 | install_requires=[ 17 | 'torch', 18 | ], 19 | classifiers=[ 20 | 'Development Status :: 4 - Beta', 21 | 'Intended Audience :: Developers', 22 | 'Topic :: Scientific/Engineering :: Artificial Intelligence', 23 | 'License :: OSI Approved :: MIT License', 24 | 'Programming Language :: Python :: 3.6', 25 | ], 26 | ) --------------------------------------------------------------------------------