lib/classification.mjs
51 |
import { BaseNeuralNetwork, } from './deep_learning';
52 |
53 | /**
54 | * Deep Learning Classification with Tensorflow
55 | * @class DeepLearningClassification
56 | * @implements {BaseNeuralNetwork}
57 | */
58 | export class DeepLearningClassification extends BaseNeuralNetwork{
59 | /**
60 | * @param {{layers:Array<Object>,compile:Object,fit:Object}} options - neural network configuration and tensorflow model hyperparameters
61 | * @param {{model:Object,tf:Object,}} properties - extra instance properties
62 | */
63 | constructor(options = {}, properties) {
64 | const config = Object.assign({
65 | layers: [],
66 | compile: {
67 | loss: 'categoricalCrossentropy',
68 | optimizer: 'adam',
69 | },
70 | fit: {
71 | epochs: 100,
72 | batchSize: 5,
73 | },
74 | }, options);
75 | super(config, properties);
76 | return this;
77 | }
78 | /**
79 | * Adds dense layers to tensorflow classification model
80 | * @override
81 | * @param {Array<Array<number>>} x_matrix - independent variables
82 | * @param {Array<Array<number>>} y_matrix - dependent variables
83 | * @param {Array<Object>} layers - model dense layer parameters
84 | */
85 | generateLayers(x_matrix, y_matrix, layers) {
86 | const xShape = this.getInputShape(x_matrix);
87 | const yShape = this.getInputShape(y_matrix);
88 | this.yShape = yShape;
89 | this.xShape = xShape;
90 | const denseLayers = [];
91 | if (layers) {
92 | denseLayers.push(...layers);
93 | } else {
94 | denseLayers.push({ units: (xShape[ 1 ] * 2), inputDim: xShape[1], activation: 'relu', });
95 | denseLayers.push({ units: yShape[ 1 ], activation: 'softmax', });
96 | }
97 | this.layers = denseLayers;
98 | denseLayers.forEach(layer => {
99 | this.model.add(this.tf.layers.dense(layer));
100 | });
101 | }
102 | }
103 |
104 |
lib/deeplearning.mjs
42 |
import { TensorScriptModelInterface, } from '@tensorscript/core';
43 |
44 | /**
45 | * Deep Learning with Tensorflow
46 | * @class BaseNeuralNetwork
47 | * @implements {TensorScriptModelInterface}
48 | */
49 | export class BaseNeuralNetwork extends TensorScriptModelInterface {
50 | /**
51 | * @param {{layers:Array<Object>,compile:Object,fit:Object}} options - neural network configuration and tensorflow model hyperparameters
52 | * @param {{model:Object,tf:Object,}} properties - extra instance properties
53 | */
54 | constructor(options = {}, properties) {
55 | const config = Object.assign({
56 | layers: [],
57 | compile: {
58 | loss: 'meanSquaredError',
59 | optimizer: 'adam',
60 | },
61 | fit: {
62 | epochs: 100,
63 | batchSize: 5,
64 | },
65 | }, options);
66 | super(config, properties);
67 | return this;
68 | }
69 | /**
70 | * Adds dense layers to tensorflow model
71 | * @abstract
72 | * @param {Array<Array<number>>} x_matrix - independent variables
73 | * @param {Array<Array<number>>} y_matrix - dependent variables
74 | * @param {Array<Object>} layers - model dense layer parameters
75 | */
76 | generateLayers(x_matrix, y_matrix, layers) {
77 | throw new ReferenceError('generateLayers method is not implemented');
78 | }
79 | /**
80 | * Asynchronously trains tensorflow model
81 | * @override
82 | * @param {Array<Array<number>>} x_matrix - independent variables
83 | * @param {Array<Array<number>>} y_matrix - dependent variables
84 | * @param {Array<Object>} layers - array of model dense layer parameters
85 | * @param {Array<Array<number>>} x_text - validation data independent variables
86 | * @param {Array<Array<number>>} y_text - validation data dependent variables
87 | * @return {Object} returns trained tensorflow model
88 | */
89 | async train(x_matrix, y_matrix, layers, x_test, y_test) {
90 | const xShape = this.getInputShape(x_matrix);
91 | const yShape = this.getInputShape(y_matrix);
92 | const xs = this.tf.tensor(x_matrix, xShape);
93 | const ys = this.tf.tensor(y_matrix, yShape);
94 | this.xShape = xShape;
95 | this.yShape = yShape;
96 | this.model = this.tf.sequential();
97 | this.generateLayers.call(this, x_matrix, y_matrix, layers || this.layers, x_test, y_test);
98 | this.model.compile(this.settings.compile);
99 | await this.model.fit(xs, ys, this.settings.fit);
100 | xs.dispose();
101 | ys.dispose();
102 | return this.model;
103 | }
104 | /**
105 | * Predicts new dependent variables
106 | * @override
107 | * @param {Array<Array<number>>|Array<number>} matrix - new test independent variables
108 | * @return {{data: Promise}} returns tensorflow prediction
109 | */
110 | calculate(input_matrix) {
111 | if (!input_matrix || Array.isArray(input_matrix)===false) throw new Error('invalid input matrix');
112 | const predictionInput = (Array.isArray(input_matrix[ 0 ]))
113 | ? input_matrix
114 | : [
115 | input_matrix,
116 | ];
117 | const predictionTensor = this.tf.tensor(predictionInput);
118 | const prediction = this.model.predict(predictionTensor);
119 | predictionTensor.dispose();
120 | return prediction;
121 | }
122 | }
123 |
124 |
lib/mlr.mjs
38 |
import { TensorScriptModelInterface, } from '@tensorscript/core';
39 |
40 | /**
41 | * Mulitple Linear Regression with Tensorflow
42 | * @class MultipleLinearRegression
43 | * @implements {TensorScriptModelInterface}
44 | */
45 | export class MultipleLinearRegression extends TensorScriptModelInterface {
46 | /**
47 | * @param {Object} options - tensorflow model hyperparameters
48 | * @param {Object} customTF - custom tensorflow module: tensorflow / tensorflow-node / tensorflow-node-gpu
49 | */
50 | constructor(options = {}, customTF) {
51 | const config = Object.assign({
52 | epochs:500,
53 | }, options);
54 | super(config, customTF);
55 | return this;
56 | }
57 | /**
58 | * Asynchronously trains tensorflow model
59 | * @override
60 | * @param {Array<Array<number>>} x_matrix - independent variables
61 | * @param {Array<Array<number>>} y_matrix - dependent variables
62 | * @return {Object} returns trained tensorflow model
63 | */
64 | async train(x_matrix, y_matrix) {
65 | const xShape = this.getInputShape(x_matrix);
66 | const yShape = this.getInputShape(y_matrix);
67 | const xs = this.tf.tensor(x_matrix, xShape);
68 | const ys = this.tf.tensor(y_matrix, yShape);
69 | this.yShape = yShape;
70 | this.xShape = xShape;
71 | this.model = this.tf.sequential();
72 | this.model.add(this.tf.layers.dense({ units: yShape[1], inputShape: [xShape[1],], }));
73 | this.model.compile({
74 | loss: 'meanSquaredError',
75 | optimizer: 'sgd',
76 | });
77 | await this.model.fit(xs, ys, this.settings);
78 | xs.dispose();
79 | ys.dispose();
80 | return this.model;
81 | }
82 | /**
83 | * Predicts new dependent variables
84 | * @override
85 | * @param {Array<Array<number>>|Array<number>} matrix - new test independent variables
86 | * @return {{data: Promise}} returns tensorflow prediction
87 | */
88 | calculate(input_matrix) {
89 | const predictionInput = (Array.isArray(input_matrix[ 0 ]))
90 | ? input_matrix
91 | : [ input_matrix, ];
92 | const predictionTensor = this.tf.tensor(predictionInput);
93 | const prediction = this.model.predict(predictionTensor);
94 | predictionTensor.dispose();
95 | return prediction;
96 | }
97 | }
98 |
99 |
lib/multiple_linear_regression.mjs
51 |
import { BaseNeuralNetwork, } from './deep_learning';
52 |
53 | /**
54 | * Mulitple Linear Regression with Tensorflow
55 | * @class MultipleLinearRegression
56 | * @implements {BaseNeuralNetwork}
57 | */
58 | export class MultipleLinearRegression extends BaseNeuralNetwork {
59 | /**
60 | * @param {{layers:Array<Object>,compile:Object,fit:Object}} options - neural network configuration and tensorflow model hyperparameters
61 | * @param {{model:Object,tf:Object,}} properties - extra instance properties
62 | */
63 | /* istanbul ignore next */
64 | constructor(options = {}, properties = {}) {
65 | const config = Object.assign({
66 | layers: [],
67 | compile: {
68 | loss: 'meanSquaredError',
69 | optimizer: 'sgd',
70 | },
71 | fit: {
72 | epochs: 500,
73 | batchSize: 5,
74 | },
75 | }, options);
76 | super(config, properties);
77 | return this;
78 | }
79 | /**
80 | * Adds dense layers to tensorflow regression model
81 | * @override
82 | * @param {Array<Array<number>>} x_matrix - independent variables
83 | * @param {Array<Array<number>>} y_matrix - dependent variables
84 | * @param {Array<Object>} layers - model dense layer parameters
85 | */
86 | generateLayers(x_matrix, y_matrix, layers) {
87 | const xShape = this.getInputShape(x_matrix);
88 | const yShape = this.getInputShape(y_matrix);
89 | const denseLayers = [];
90 | if (layers) {
91 | denseLayers.push(...layers);
92 | } else {
93 | denseLayers.push({ units: yShape[1], inputShape: [xShape[1],], });
94 | }
95 | this.layers = denseLayers;
96 | denseLayers.forEach(layer => {
97 | this.model.add(this.tf.layers.dense(layer));
98 | });
99 | }
100 | }
101 |
102 |
29 |
30 | 
