nntsreinforcement/lib/network/neuron.js

A Neural Network for Reinforcement Learning in TypeScript

"use strict";
var __spreadArray = (this && this.__spreadArray) || function (to, from, pack) {
    if (pack || arguments.length === 2) for (var i = 0, l = from.length, ar; i < l; i++) {
        if (ar || !(i in from)) {
            if (!ar) ar = Array.prototype.slice.call(from, 0, i);
            ar[i] = from[i];
        }
    }
    return to.concat(ar || Array.prototype.slice.call(from));
};
Object.defineProperty(exports, "__esModule", { value: true });
exports.Neuron = void 0;
var Neuron = /** @class */ (function () {
    function Neuron(bias, weights, type) {
        if (bias === void 0) { bias = Math.random() * 2 - 1; }
        if (type === void 0) { type = 'Type not defined'; }
        this.type = type;
        this.activation = 0;
        this.bias = 0;
        this.weights = [];
        this.prevNeurons = [];
        this.bias = bias;
        if (weights != undefined) {
            this.weights = weights;
        }
    }
    Neuron.prototype.initRandomConnectionsFromLayer = function (prevLayer, connections) {
        this.prevNeurons = prevLayer.neurons;
        for (var _i = 0, _a = prevLayer.neurons; _i < _a.length; _i++) {
            var neuron = _a[_i];
            this.weights.push(Math.random() * 2 - 1);
        }
    };
    Neuron.prototype.compute = function (activations) {
        var sum = -this.bias;
        for (var i = 0; i < activations.length; i++) {
            sum += this.weights[i] * activations[i];
        }
        this.activation = this.relu(sum);
        return this.activation;
    };
    Neuron.prototype.binaryActivation = function (sum) {
        return sum < this.bias ? 0 : 1;
    };
    Neuron.prototype.sigmoid = function (a) {
        return 1 / (1 + Math.pow(Math.E, -a));
    };
    Neuron.prototype.relu = function (a) {
        return a > 0 ? a : 0;
    };
    Neuron.prototype.mutate = function (rate) {
        this.bias = this.linearInterpolation(this.bias, Math.random() * 2 - 1, rate);
        for (var i = 0; i < this.weights.length; i++) {
            this.weights[i] = this.linearInterpolation(this.weights[i], Math.random() * 2 - 1, rate);
        }
    };
    Neuron.prototype.deepCopy = function () {
        var newWeights = __spreadArray([], this.weights, true);
        return new Neuron(this.bias, newWeights, this.type);
    };
    Neuron.prototype.linearInterpolation = function (a, b, d) {
        return a + (b - a) * d;
    };
    return Neuron;
}());
exports.Neuron = Neuron;