Vsevolod Rodionov
34c93cb718
that is related to https://github.com/cazala/synaptic/issues/137, we're using fn.toString() and bunch of regex-es to create a worker for now (I know, it's bad, still we already have a fix, but it is breaking some of our actually-interal, but overly-exposed APIs, so it goes to v2), and we need to either have one version minified by us or it will be minified by CDNs.
2848 linhas
91 KiB
JavaScript
2848 linhas
91 KiB
JavaScript
/*!
|
|
* The MIT License (MIT)
|
|
*
|
|
* Copyright (c) 2016 Juan Cazala - juancazala.com
|
|
*
|
|
* Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
* of this software and associated documentation files (the "Software"), to deal
|
|
* in the Software without restriction, including without limitation the rights
|
|
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
* copies of the Software, and to permit persons to whom the Software is
|
|
* furnished to do so, subject to the following conditions:
|
|
*
|
|
* The above copyright notice and this permission notice shall be included in
|
|
* all copies or substantial portions of the Software.
|
|
*
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
|
* THE SOFTWARE
|
|
*
|
|
*
|
|
*
|
|
* ********************************************************************************************
|
|
* SYNAPTIC (v1.0.8)
|
|
* ********************************************************************************************
|
|
*
|
|
* Synaptic is a javascript neural network library for node.js and the browser, its generalized
|
|
* algorithm is architecture-free, so you can build and train basically any type of first order
|
|
* or even second order neural network architectures.
|
|
*
|
|
* http://en.wikipedia.org/wiki/Recurrent_neural_network#Second_Order_Recurrent_Neural_Network
|
|
*
|
|
* The library includes a few built-in architectures like multilayer perceptrons, multilayer
|
|
* long-short term memory networks (LSTM) or liquid state machines, and a trainer capable of
|
|
* training any given network, and includes built-in training tasks/tests like solving an XOR,
|
|
* passing a Distracted Sequence Recall test or an Embeded Reber Grammar test.
|
|
*
|
|
* The algorithm implemented by this library has been taken from Derek D. Monner's paper:
|
|
*
|
|
*
|
|
* A generalized LSTM-like training algorithm for second-order recurrent neural networks
|
|
* http://www.overcomplete.net/papers/nn2012.pdf
|
|
*
|
|
* There are references to the equations in that paper commented through the source code.
|
|
*
|
|
*/
|
|
/******/ (function(modules) { // webpackBootstrap
|
|
/******/ // The module cache
|
|
/******/ var installedModules = {};
|
|
|
|
/******/ // The require function
|
|
/******/ function __webpack_require__(moduleId) {
|
|
|
|
/******/ // Check if module is in cache
|
|
/******/ if(installedModules[moduleId])
|
|
/******/ return installedModules[moduleId].exports;
|
|
|
|
/******/ // Create a new module (and put it into the cache)
|
|
/******/ var module = installedModules[moduleId] = {
|
|
/******/ exports: {},
|
|
/******/ id: moduleId,
|
|
/******/ loaded: false
|
|
/******/ };
|
|
|
|
/******/ // Execute the module function
|
|
/******/ modules[moduleId].call(module.exports, module, module.exports, __webpack_require__);
|
|
|
|
/******/ // Flag the module as loaded
|
|
/******/ module.loaded = true;
|
|
|
|
/******/ // Return the exports of the module
|
|
/******/ return module.exports;
|
|
/******/ }
|
|
|
|
|
|
/******/ // expose the modules object (__webpack_modules__)
|
|
/******/ __webpack_require__.m = modules;
|
|
|
|
/******/ // expose the module cache
|
|
/******/ __webpack_require__.c = installedModules;
|
|
|
|
/******/ // __webpack_public_path__
|
|
/******/ __webpack_require__.p = "";
|
|
|
|
/******/ // Load entry module and return exports
|
|
/******/ return __webpack_require__(0);
|
|
/******/ })
|
|
/************************************************************************/
|
|
/******/ ([
|
|
/* 0 */
|
|
/***/ function(module, exports, __webpack_require__) {
|
|
|
|
var __WEBPACK_AMD_DEFINE_ARRAY__, __WEBPACK_AMD_DEFINE_RESULT__;var Synaptic = {
|
|
Neuron: __webpack_require__(1),
|
|
Layer: __webpack_require__(3),
|
|
Network: __webpack_require__(4),
|
|
Trainer: __webpack_require__(5),
|
|
Architect: __webpack_require__(6)
|
|
};
|
|
|
|
// CommonJS & AMD
|
|
if (true)
|
|
{
|
|
!(__WEBPACK_AMD_DEFINE_ARRAY__ = [], __WEBPACK_AMD_DEFINE_RESULT__ = function(){ return Synaptic }.apply(exports, __WEBPACK_AMD_DEFINE_ARRAY__), __WEBPACK_AMD_DEFINE_RESULT__ !== undefined && (module.exports = __WEBPACK_AMD_DEFINE_RESULT__));
|
|
}
|
|
|
|
// Node.js
|
|
if (typeof module !== 'undefined' && module.exports)
|
|
{
|
|
module.exports = Synaptic;
|
|
}
|
|
|
|
// Browser
|
|
if (typeof window == 'object')
|
|
{
|
|
(function(){
|
|
var oldSynaptic = window['synaptic'];
|
|
Synaptic.ninja = function(){
|
|
window['synaptic'] = oldSynaptic;
|
|
return Synaptic;
|
|
};
|
|
})();
|
|
|
|
window['synaptic'] = Synaptic;
|
|
}
|
|
|
|
|
|
/***/ },
|
|
/* 1 */
|
|
/***/ function(module, exports, __webpack_require__) {
|
|
|
|
/* WEBPACK VAR INJECTION */(function(module) {// export
|
|
if (module) module.exports = Neuron;
|
|
|
|
/******************************************************************************************
|
|
NEURON
|
|
*******************************************************************************************/
|
|
|
|
function Neuron() {
|
|
this.ID = Neuron.uid();
|
|
this.label = null;
|
|
this.connections = {
|
|
inputs: {},
|
|
projected: {},
|
|
gated: {}
|
|
};
|
|
this.error = {
|
|
responsibility: 0,
|
|
projected: 0,
|
|
gated: 0
|
|
};
|
|
this.trace = {
|
|
elegibility: {},
|
|
extended: {},
|
|
influences: {}
|
|
};
|
|
this.state = 0;
|
|
this.old = 0;
|
|
this.activation = 0;
|
|
this.selfconnection = new Neuron.connection(this, this, 0); // weight = 0 -> not connected
|
|
this.squash = Neuron.squash.LOGISTIC;
|
|
this.neighboors = {};
|
|
this.bias = Math.random() * .2 - .1;
|
|
}
|
|
|
|
Neuron.prototype = {
|
|
|
|
// activate the neuron
|
|
activate: function(input) {
|
|
// activation from enviroment (for input neurons)
|
|
if (typeof input != 'undefined') {
|
|
this.activation = input;
|
|
this.derivative = 0;
|
|
this.bias = 0;
|
|
return this.activation;
|
|
}
|
|
|
|
// old state
|
|
this.old = this.state;
|
|
|
|
// eq. 15
|
|
this.state = this.selfconnection.gain * this.selfconnection.weight *
|
|
this.state + this.bias;
|
|
|
|
for (var i in this.connections.inputs) {
|
|
var input = this.connections.inputs[i];
|
|
this.state += input.from.activation * input.weight * input.gain;
|
|
}
|
|
|
|
// eq. 16
|
|
this.activation = this.squash(this.state);
|
|
|
|
// f'(s)
|
|
this.derivative = this.squash(this.state, true);
|
|
|
|
// update traces
|
|
var influences = [];
|
|
for (var id in this.trace.extended) {
|
|
// extended elegibility trace
|
|
var neuron = this.neighboors[id];
|
|
|
|
// if gated neuron's selfconnection is gated by this unit, the influence keeps track of the neuron's old state
|
|
var influence = neuron.selfconnection.gater == this ? neuron.old : 0;
|
|
|
|
// index runs over all the incoming connections to the gated neuron that are gated by this unit
|
|
for (var incoming in this.trace.influences[neuron.ID]) { // captures the effect that has an input connection to this unit, on a neuron that is gated by this unit
|
|
influence += this.trace.influences[neuron.ID][incoming].weight *
|
|
this.trace.influences[neuron.ID][incoming].from.activation;
|
|
}
|
|
influences[neuron.ID] = influence;
|
|
}
|
|
|
|
for (var i in this.connections.inputs) {
|
|
var input = this.connections.inputs[i];
|
|
|
|
// elegibility trace - Eq. 17
|
|
this.trace.elegibility[input.ID] = this.selfconnection.gain * this.selfconnection
|
|
.weight * this.trace.elegibility[input.ID] + input.gain * input.from
|
|
.activation;
|
|
|
|
for (var id in this.trace.extended) {
|
|
// extended elegibility trace
|
|
var xtrace = this.trace.extended[id];
|
|
var neuron = this.neighboors[id];
|
|
var influence = influences[neuron.ID];
|
|
|
|
// eq. 18
|
|
xtrace[input.ID] = neuron.selfconnection.gain * neuron.selfconnection
|
|
.weight * xtrace[input.ID] + this.derivative * this.trace.elegibility[
|
|
input.ID] * influence;
|
|
}
|
|
}
|
|
|
|
// update gated connection's gains
|
|
for (var connection in this.connections.gated) {
|
|
this.connections.gated[connection].gain = this.activation;
|
|
}
|
|
|
|
return this.activation;
|
|
},
|
|
|
|
// back-propagate the error
|
|
propagate: function(rate, target) {
|
|
// error accumulator
|
|
var error = 0;
|
|
|
|
// whether or not this neuron is in the output layer
|
|
var isOutput = typeof target != 'undefined';
|
|
|
|
// output neurons get their error from the enviroment
|
|
if (isOutput)
|
|
this.error.responsibility = this.error.projected = target - this.activation; // Eq. 10
|
|
|
|
else // the rest of the neuron compute their error responsibilities by backpropagation
|
|
{
|
|
// error responsibilities from all the connections projected from this neuron
|
|
for (var id in this.connections.projected) {
|
|
var connection = this.connections.projected[id];
|
|
var neuron = connection.to;
|
|
// Eq. 21
|
|
error += neuron.error.responsibility * connection.gain * connection.weight;
|
|
}
|
|
|
|
// projected error responsibility
|
|
this.error.projected = this.derivative * error;
|
|
|
|
error = 0;
|
|
// error responsibilities from all the connections gated by this neuron
|
|
for (var id in this.trace.extended) {
|
|
var neuron = this.neighboors[id]; // gated neuron
|
|
var influence = neuron.selfconnection.gater == this ? neuron.old : 0; // if gated neuron's selfconnection is gated by this neuron
|
|
|
|
// index runs over all the connections to the gated neuron that are gated by this neuron
|
|
for (var input in this.trace.influences[id]) { // captures the effect that the input connection of this neuron have, on a neuron which its input/s is/are gated by this neuron
|
|
influence += this.trace.influences[id][input].weight * this.trace.influences[
|
|
neuron.ID][input].from.activation;
|
|
}
|
|
// eq. 22
|
|
error += neuron.error.responsibility * influence;
|
|
}
|
|
|
|
// gated error responsibility
|
|
this.error.gated = this.derivative * error;
|
|
|
|
// error responsibility - Eq. 23
|
|
this.error.responsibility = this.error.projected + this.error.gated;
|
|
}
|
|
|
|
// learning rate
|
|
rate = rate || .1;
|
|
|
|
// adjust all the neuron's incoming connections
|
|
for (var id in this.connections.inputs) {
|
|
var input = this.connections.inputs[id];
|
|
|
|
// Eq. 24
|
|
var gradient = this.error.projected * this.trace.elegibility[input.ID];
|
|
for (var id in this.trace.extended) {
|
|
var neuron = this.neighboors[id];
|
|
gradient += neuron.error.responsibility * this.trace.extended[
|
|
neuron.ID][input.ID];
|
|
}
|
|
input.weight += rate * gradient; // adjust weights - aka learn
|
|
}
|
|
|
|
// adjust bias
|
|
this.bias += rate * this.error.responsibility;
|
|
},
|
|
|
|
project: function(neuron, weight) {
|
|
// self-connection
|
|
if (neuron == this) {
|
|
this.selfconnection.weight = 1;
|
|
return this.selfconnection;
|
|
}
|
|
|
|
// check if connection already exists
|
|
var connected = this.connected(neuron);
|
|
if (connected && connected.type == "projected") {
|
|
// update connection
|
|
if (typeof weight != 'undefined')
|
|
connected.connection.weight = weight;
|
|
// return existing connection
|
|
return connected.connection;
|
|
} else {
|
|
// create a new connection
|
|
var connection = new Neuron.connection(this, neuron, weight);
|
|
}
|
|
|
|
// reference all the connections and traces
|
|
this.connections.projected[connection.ID] = connection;
|
|
this.neighboors[neuron.ID] = neuron;
|
|
neuron.connections.inputs[connection.ID] = connection;
|
|
neuron.trace.elegibility[connection.ID] = 0;
|
|
|
|
for (var id in neuron.trace.extended) {
|
|
var trace = neuron.trace.extended[id];
|
|
trace[connection.ID] = 0;
|
|
}
|
|
|
|
return connection;
|
|
},
|
|
|
|
gate: function(connection) {
|
|
// add connection to gated list
|
|
this.connections.gated[connection.ID] = connection;
|
|
|
|
var neuron = connection.to;
|
|
if (!(neuron.ID in this.trace.extended)) {
|
|
// extended trace
|
|
this.neighboors[neuron.ID] = neuron;
|
|
var xtrace = this.trace.extended[neuron.ID] = {};
|
|
for (var id in this.connections.inputs) {
|
|
var input = this.connections.inputs[id];
|
|
xtrace[input.ID] = 0;
|
|
}
|
|
}
|
|
|
|
// keep track
|
|
if (neuron.ID in this.trace.influences)
|
|
this.trace.influences[neuron.ID].push(connection);
|
|
else
|
|
this.trace.influences[neuron.ID] = [connection];
|
|
|
|
// set gater
|
|
connection.gater = this;
|
|
},
|
|
|
|
// returns true or false whether the neuron is self-connected or not
|
|
selfconnected: function() {
|
|
return this.selfconnection.weight !== 0;
|
|
},
|
|
|
|
// returns true or false whether the neuron is connected to another neuron (parameter)
|
|
connected: function(neuron) {
|
|
var result = {
|
|
type: null,
|
|
connection: false
|
|
};
|
|
|
|
if (this == neuron) {
|
|
if (this.selfconnected()) {
|
|
result.type = 'selfconnection';
|
|
result.connection = this.selfconnection;
|
|
return result;
|
|
} else
|
|
return false;
|
|
}
|
|
|
|
for (var type in this.connections) {
|
|
for (var connection in this.connections[type]) {
|
|
var connection = this.connections[type][connection];
|
|
if (connection.to == neuron) {
|
|
result.type = type;
|
|
result.connection = connection;
|
|
return result;
|
|
} else if (connection.from == neuron) {
|
|
result.type = type;
|
|
result.connection = connection;
|
|
return result;
|
|
}
|
|
}
|
|
}
|
|
|
|
return false;
|
|
},
|
|
|
|
// clears all the traces (the neuron forgets it's context, but the connections remain intact)
|
|
clear: function() {
|
|
|
|
for (var trace in this.trace.elegibility)
|
|
this.trace.elegibility[trace] = 0;
|
|
|
|
for (var trace in this.trace.extended)
|
|
for (var extended in this.trace.extended[trace])
|
|
this.trace.extended[trace][extended] = 0;
|
|
|
|
this.error.responsibility = this.error.projected = this.error.gated = 0;
|
|
},
|
|
|
|
// all the connections are randomized and the traces are cleared
|
|
reset: function() {
|
|
this.clear();
|
|
|
|
for (var type in this.connections)
|
|
for (var connection in this.connections[type])
|
|
this.connections[type][connection].weight = Math.random() * .2 - .1;
|
|
this.bias = Math.random() * .2 - .1;
|
|
|
|
this.old = this.state = this.activation = 0;
|
|
},
|
|
|
|
// hardcodes the behaviour of the neuron into an optimized function
|
|
optimize: function(optimized, layer) {
|
|
|
|
optimized = optimized || {};
|
|
var store_activation = [];
|
|
var store_trace = [];
|
|
var store_propagation = [];
|
|
var varID = optimized.memory || 0;
|
|
var neurons = optimized.neurons || 1;
|
|
var inputs = optimized.inputs || [];
|
|
var targets = optimized.targets || [];
|
|
var outputs = optimized.outputs || [];
|
|
var variables = optimized.variables || {};
|
|
var activation_sentences = optimized.activation_sentences || [];
|
|
var trace_sentences = optimized.trace_sentences || [];
|
|
var propagation_sentences = optimized.propagation_sentences || [];
|
|
var layers = optimized.layers || { __count: 0, __neuron: 0 };
|
|
|
|
// allocate sentences
|
|
var allocate = function(store){
|
|
var allocated = layer in layers && store[layers.__count];
|
|
if (!allocated)
|
|
{
|
|
layers.__count = store.push([]) - 1;
|
|
layers[layer] = layers.__count;
|
|
}
|
|
};
|
|
allocate(activation_sentences);
|
|
allocate(trace_sentences);
|
|
allocate(propagation_sentences);
|
|
var currentLayer = layers.__count;
|
|
|
|
// get/reserve space in memory by creating a unique ID for a variablel
|
|
var getVar = function() {
|
|
var args = Array.prototype.slice.call(arguments);
|
|
|
|
if (args.length == 1) {
|
|
if (args[0] == 'target') {
|
|
var id = 'target_' + targets.length;
|
|
targets.push(varID);
|
|
} else
|
|
var id = args[0];
|
|
if (id in variables)
|
|
return variables[id];
|
|
return variables[id] = {
|
|
value: 0,
|
|
id: varID++
|
|
};
|
|
} else {
|
|
var extended = args.length > 2;
|
|
if (extended)
|
|
var value = args.pop();
|
|
|
|
var unit = args.shift();
|
|
var prop = args.pop();
|
|
|
|
if (!extended)
|
|
var value = unit[prop];
|
|
|
|
var id = prop + '_';
|
|
for (var property in args)
|
|
id += args[property] + '_';
|
|
id += unit.ID;
|
|
if (id in variables)
|
|
return variables[id];
|
|
|
|
return variables[id] = {
|
|
value: value,
|
|
id: varID++
|
|
};
|
|
}
|
|
};
|
|
|
|
// build sentence
|
|
var buildSentence = function() {
|
|
var args = Array.prototype.slice.call(arguments);
|
|
var store = args.pop();
|
|
var sentence = "";
|
|
for (var i in args)
|
|
if (typeof args[i] == 'string')
|
|
sentence += args[i];
|
|
else
|
|
sentence += 'F[' + args[i].id + ']';
|
|
|
|
store.push(sentence + ';');
|
|
};
|
|
|
|
// helper to check if an object is empty
|
|
var isEmpty = function(obj) {
|
|
for (var prop in obj) {
|
|
if (obj.hasOwnProperty(prop))
|
|
return false;
|
|
}
|
|
return true;
|
|
};
|
|
|
|
// characteristics of the neuron
|
|
var noProjections = isEmpty(this.connections.projected);
|
|
var noGates = isEmpty(this.connections.gated);
|
|
var isInput = layer == 'input' ? true : isEmpty(this.connections.inputs);
|
|
var isOutput = layer == 'output' ? true : noProjections && noGates;
|
|
|
|
// optimize neuron's behaviour
|
|
var rate = getVar('rate');
|
|
var activation = getVar(this, 'activation');
|
|
if (isInput)
|
|
inputs.push(activation.id);
|
|
else {
|
|
activation_sentences[currentLayer].push(store_activation);
|
|
trace_sentences[currentLayer].push(store_trace);
|
|
propagation_sentences[currentLayer].push(store_propagation);
|
|
var old = getVar(this, 'old');
|
|
var state = getVar(this, 'state');
|
|
var bias = getVar(this, 'bias');
|
|
if (this.selfconnection.gater)
|
|
var self_gain = getVar(this.selfconnection, 'gain');
|
|
if (this.selfconnected())
|
|
var self_weight = getVar(this.selfconnection, 'weight');
|
|
buildSentence(old, ' = ', state, store_activation);
|
|
if (this.selfconnected())
|
|
if (this.selfconnection.gater)
|
|
buildSentence(state, ' = ', self_gain, ' * ', self_weight, ' * ',
|
|
state, ' + ', bias, store_activation);
|
|
else
|
|
buildSentence(state, ' = ', self_weight, ' * ', state, ' + ',
|
|
bias, store_activation);
|
|
else
|
|
buildSentence(state, ' = ', bias, store_activation);
|
|
for (var i in this.connections.inputs) {
|
|
var input = this.connections.inputs[i];
|
|
var input_activation = getVar(input.from, 'activation');
|
|
var input_weight = getVar(input, 'weight');
|
|
if (input.gater)
|
|
var input_gain = getVar(input, 'gain');
|
|
if (this.connections.inputs[i].gater)
|
|
buildSentence(state, ' += ', input_activation, ' * ',
|
|
input_weight, ' * ', input_gain, store_activation);
|
|
else
|
|
buildSentence(state, ' += ', input_activation, ' * ',
|
|
input_weight, store_activation);
|
|
}
|
|
var derivative = getVar(this, 'derivative');
|
|
switch (this.squash) {
|
|
case Neuron.squash.LOGISTIC:
|
|
buildSentence(activation, ' = (1 / (1 + Math.exp(-', state, ')))',
|
|
store_activation);
|
|
buildSentence(derivative, ' = ', activation, ' * (1 - ',
|
|
activation, ')', store_activation);
|
|
break;
|
|
case Neuron.squash.TANH:
|
|
var eP = getVar('aux');
|
|
var eN = getVar('aux_2');
|
|
buildSentence(eP, ' = Math.exp(', state, ')', store_activation);
|
|
buildSentence(eN, ' = 1 / ', eP, store_activation);
|
|
buildSentence(activation, ' = (', eP, ' - ', eN, ') / (', eP, ' + ', eN, ')', store_activation);
|
|
buildSentence(derivative, ' = 1 - (', activation, ' * ', activation, ')', store_activation);
|
|
break;
|
|
case Neuron.squash.IDENTITY:
|
|
buildSentence(activation, ' = ', state, store_activation);
|
|
buildSentence(derivative, ' = 1', store_activation);
|
|
break;
|
|
case Neuron.squash.HLIM:
|
|
buildSentence(activation, ' = +(', state, ' > 0)', store_activation);
|
|
buildSentence(derivative, ' = 1', store_activation);
|
|
case Neuron.squash.RELU:
|
|
buildSentence(activation, ' = ', state, ' > 0 ? ', state, ' : 0', store_activation);
|
|
buildSentence(derivative, ' = ', state, ' > 0 ? 1 : 0', store_activation);
|
|
break;
|
|
}
|
|
|
|
for (var id in this.trace.extended) {
|
|
// calculate extended elegibility traces in advance
|
|
|
|
var neuron = this.neighboors[id];
|
|
var influence = getVar('influences[' + neuron.ID + ']');
|
|
var neuron_old = getVar(neuron, 'old');
|
|
var initialized = false;
|
|
if (neuron.selfconnection.gater == this)
|
|
{
|
|
buildSentence(influence, ' = ', neuron_old, store_trace);
|
|
initialized = true;
|
|
}
|
|
for (var incoming in this.trace.influences[neuron.ID]) {
|
|
var incoming_weight = getVar(this.trace.influences[neuron.ID]
|
|
[incoming], 'weight');
|
|
var incoming_activation = getVar(this.trace.influences[neuron.ID]
|
|
[incoming].from, 'activation');
|
|
|
|
if (initialized)
|
|
buildSentence(influence, ' += ', incoming_weight, ' * ', incoming_activation, store_trace);
|
|
else {
|
|
buildSentence(influence, ' = ', incoming_weight, ' * ', incoming_activation, store_trace);
|
|
initialized = true;
|
|
}
|
|
}
|
|
}
|
|
|
|
for (var i in this.connections.inputs) {
|
|
var input = this.connections.inputs[i];
|
|
if (input.gater)
|
|
var input_gain = getVar(input, 'gain');
|
|
var input_activation = getVar(input.from, 'activation');
|
|
var trace = getVar(this, 'trace', 'elegibility', input.ID, this.trace
|
|
.elegibility[input.ID]);
|
|
if (this.selfconnected()) {
|
|
if (this.selfconnection.gater) {
|
|
if (input.gater)
|
|
buildSentence(trace, ' = ', self_gain, ' * ', self_weight,
|
|
' * ', trace, ' + ', input_gain, ' * ', input_activation,
|
|
store_trace);
|
|
else
|
|
buildSentence(trace, ' = ', self_gain, ' * ', self_weight,
|
|
' * ', trace, ' + ', input_activation, store_trace);
|
|
} else {
|
|
if (input.gater)
|
|
buildSentence(trace, ' = ', self_weight, ' * ', trace, ' + ',
|
|
input_gain, ' * ', input_activation, store_trace);
|
|
else
|
|
buildSentence(trace, ' = ', self_weight, ' * ', trace, ' + ',
|
|
input_activation, store_trace);
|
|
}
|
|
} else {
|
|
if (input.gater)
|
|
buildSentence(trace, ' = ', input_gain, ' * ', input_activation,
|
|
store_trace);
|
|
else
|
|
buildSentence(trace, ' = ', input_activation, store_trace);
|
|
}
|
|
for (var id in this.trace.extended) {
|
|
// extended elegibility trace
|
|
var neuron = this.neighboors[id];
|
|
var influence = getVar('influences[' + neuron.ID + ']');
|
|
|
|
var trace = getVar(this, 'trace', 'elegibility', input.ID, this.trace
|
|
.elegibility[input.ID]);
|
|
var xtrace = getVar(this, 'trace', 'extended', neuron.ID, input.ID,
|
|
this.trace.extended[neuron.ID][input.ID]);
|
|
if (neuron.selfconnected())
|
|
var neuron_self_weight = getVar(neuron.selfconnection, 'weight');
|
|
if (neuron.selfconnection.gater)
|
|
var neuron_self_gain = getVar(neuron.selfconnection, 'gain');
|
|
if (neuron.selfconnected())
|
|
if (neuron.selfconnection.gater)
|
|
buildSentence(xtrace, ' = ', neuron_self_gain, ' * ',
|
|
neuron_self_weight, ' * ', xtrace, ' + ', derivative, ' * ',
|
|
trace, ' * ', influence, store_trace);
|
|
else
|
|
buildSentence(xtrace, ' = ', neuron_self_weight, ' * ',
|
|
xtrace, ' + ', derivative, ' * ', trace, ' * ',
|
|
influence, store_trace);
|
|
else
|
|
buildSentence(xtrace, ' = ', derivative, ' * ', trace, ' * ',
|
|
influence, store_trace);
|
|
}
|
|
}
|
|
for (var connection in this.connections.gated) {
|
|
var gated_gain = getVar(this.connections.gated[connection], 'gain');
|
|
buildSentence(gated_gain, ' = ', activation, store_activation);
|
|
}
|
|
}
|
|
if (!isInput) {
|
|
var responsibility = getVar(this, 'error', 'responsibility', this.error
|
|
.responsibility);
|
|
if (isOutput) {
|
|
var target = getVar('target');
|
|
buildSentence(responsibility, ' = ', target, ' - ', activation,
|
|
store_propagation);
|
|
for (var id in this.connections.inputs) {
|
|
var input = this.connections.inputs[id];
|
|
var trace = getVar(this, 'trace', 'elegibility', input.ID, this.trace
|
|
.elegibility[input.ID]);
|
|
var input_weight = getVar(input, 'weight');
|
|
buildSentence(input_weight, ' += ', rate, ' * (', responsibility,
|
|
' * ', trace, ')', store_propagation);
|
|
}
|
|
outputs.push(activation.id);
|
|
} else {
|
|
if (!noProjections && !noGates) {
|
|
var error = getVar('aux');
|
|
for (var id in this.connections.projected) {
|
|
var connection = this.connections.projected[id];
|
|
var neuron = connection.to;
|
|
var connection_weight = getVar(connection, 'weight');
|
|
var neuron_responsibility = getVar(neuron, 'error',
|
|
'responsibility', neuron.error.responsibility);
|
|
if (connection.gater) {
|
|
var connection_gain = getVar(connection, 'gain');
|
|
buildSentence(error, ' += ', neuron_responsibility, ' * ',
|
|
connection_gain, ' * ', connection_weight,
|
|
store_propagation);
|
|
} else
|
|
buildSentence(error, ' += ', neuron_responsibility, ' * ',
|
|
connection_weight, store_propagation);
|
|
}
|
|
var projected = getVar(this, 'error', 'projected', this.error.projected);
|
|
buildSentence(projected, ' = ', derivative, ' * ', error,
|
|
store_propagation);
|
|
buildSentence(error, ' = 0', store_propagation);
|
|
for (var id in this.trace.extended) {
|
|
var neuron = this.neighboors[id];
|
|
var influence = getVar('aux_2');
|
|
var neuron_old = getVar(neuron, 'old');
|
|
if (neuron.selfconnection.gater == this)
|
|
buildSentence(influence, ' = ', neuron_old, store_propagation);
|
|
else
|
|
buildSentence(influence, ' = 0', store_propagation);
|
|
for (var input in this.trace.influences[neuron.ID]) {
|
|
var connection = this.trace.influences[neuron.ID][input];
|
|
var connection_weight = getVar(connection, 'weight');
|
|
var neuron_activation = getVar(connection.from, 'activation');
|
|
buildSentence(influence, ' += ', connection_weight, ' * ',
|
|
neuron_activation, store_propagation);
|
|
}
|
|
var neuron_responsibility = getVar(neuron, 'error',
|
|
'responsibility', neuron.error.responsibility);
|
|
buildSentence(error, ' += ', neuron_responsibility, ' * ',
|
|
influence, store_propagation);
|
|
}
|
|
var gated = getVar(this, 'error', 'gated', this.error.gated);
|
|
buildSentence(gated, ' = ', derivative, ' * ', error,
|
|
store_propagation);
|
|
buildSentence(responsibility, ' = ', projected, ' + ', gated,
|
|
store_propagation);
|
|
for (var id in this.connections.inputs) {
|
|
var input = this.connections.inputs[id];
|
|
var gradient = getVar('aux');
|
|
var trace = getVar(this, 'trace', 'elegibility', input.ID, this
|
|
.trace.elegibility[input.ID]);
|
|
buildSentence(gradient, ' = ', projected, ' * ', trace,
|
|
store_propagation);
|
|
for (var id in this.trace.extended) {
|
|
var neuron = this.neighboors[id];
|
|
var neuron_responsibility = getVar(neuron, 'error',
|
|
'responsibility', neuron.error.responsibility);
|
|
var xtrace = getVar(this, 'trace', 'extended', neuron.ID,
|
|
input.ID, this.trace.extended[neuron.ID][input.ID]);
|
|
buildSentence(gradient, ' += ', neuron_responsibility, ' * ',
|
|
xtrace, store_propagation);
|
|
}
|
|
var input_weight = getVar(input, 'weight');
|
|
buildSentence(input_weight, ' += ', rate, ' * ', gradient,
|
|
store_propagation);
|
|
}
|
|
|
|
} else if (noGates) {
|
|
buildSentence(responsibility, ' = 0', store_propagation);
|
|
for (var id in this.connections.projected) {
|
|
var connection = this.connections.projected[id];
|
|
var neuron = connection.to;
|
|
var connection_weight = getVar(connection, 'weight');
|
|
var neuron_responsibility = getVar(neuron, 'error',
|
|
'responsibility', neuron.error.responsibility);
|
|
if (connection.gater) {
|
|
var connection_gain = getVar(connection, 'gain');
|
|
buildSentence(responsibility, ' += ', neuron_responsibility,
|
|
' * ', connection_gain, ' * ', connection_weight,
|
|
store_propagation);
|
|
} else
|
|
buildSentence(responsibility, ' += ', neuron_responsibility,
|
|
' * ', connection_weight, store_propagation);
|
|
}
|
|
buildSentence(responsibility, ' *= ', derivative,
|
|
store_propagation);
|
|
for (var id in this.connections.inputs) {
|
|
var input = this.connections.inputs[id];
|
|
var trace = getVar(this, 'trace', 'elegibility', input.ID, this
|
|
.trace.elegibility[input.ID]);
|
|
var input_weight = getVar(input, 'weight');
|
|
buildSentence(input_weight, ' += ', rate, ' * (',
|
|
responsibility, ' * ', trace, ')', store_propagation);
|
|
}
|
|
} else if (noProjections) {
|
|
buildSentence(responsibility, ' = 0', store_propagation);
|
|
for (var id in this.trace.extended) {
|
|
var neuron = this.neighboors[id];
|
|
var influence = getVar('aux');
|
|
var neuron_old = getVar(neuron, 'old');
|
|
if (neuron.selfconnection.gater == this)
|
|
buildSentence(influence, ' = ', neuron_old, store_propagation);
|
|
else
|
|
buildSentence(influence, ' = 0', store_propagation);
|
|
for (var input in this.trace.influences[neuron.ID]) {
|
|
var connection = this.trace.influences[neuron.ID][input];
|
|
var connection_weight = getVar(connection, 'weight');
|
|
var neuron_activation = getVar(connection.from, 'activation');
|
|
buildSentence(influence, ' += ', connection_weight, ' * ',
|
|
neuron_activation, store_propagation);
|
|
}
|
|
var neuron_responsibility = getVar(neuron, 'error',
|
|
'responsibility', neuron.error.responsibility);
|
|
buildSentence(responsibility, ' += ', neuron_responsibility,
|
|
' * ', influence, store_propagation);
|
|
}
|
|
buildSentence(responsibility, ' *= ', derivative,
|
|
store_propagation);
|
|
for (var id in this.connections.inputs) {
|
|
var input = this.connections.inputs[id];
|
|
var gradient = getVar('aux');
|
|
buildSentence(gradient, ' = 0', store_propagation);
|
|
for (var id in this.trace.extended) {
|
|
var neuron = this.neighboors[id];
|
|
var neuron_responsibility = getVar(neuron, 'error',
|
|
'responsibility', neuron.error.responsibility);
|
|
var xtrace = getVar(this, 'trace', 'extended', neuron.ID,
|
|
input.ID, this.trace.extended[neuron.ID][input.ID]);
|
|
buildSentence(gradient, ' += ', neuron_responsibility, ' * ',
|
|
xtrace, store_propagation);
|
|
}
|
|
var input_weight = getVar(input, 'weight');
|
|
buildSentence(input_weight, ' += ', rate, ' * ', gradient,
|
|
store_propagation);
|
|
}
|
|
}
|
|
}
|
|
buildSentence(bias, ' += ', rate, ' * ', responsibility,
|
|
store_propagation);
|
|
}
|
|
return {
|
|
memory: varID,
|
|
neurons: neurons + 1,
|
|
inputs: inputs,
|
|
outputs: outputs,
|
|
targets: targets,
|
|
variables: variables,
|
|
activation_sentences: activation_sentences,
|
|
trace_sentences: trace_sentences,
|
|
propagation_sentences: propagation_sentences,
|
|
layers: layers
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
// represents a connection between two neurons
|
|
Neuron.connection = function Connection(from, to, weight) {
|
|
|
|
if (!from || !to)
|
|
throw new Error("Connection Error: Invalid neurons");
|
|
|
|
this.ID = Neuron.connection.uid();
|
|
this.from = from;
|
|
this.to = to;
|
|
this.weight = typeof weight == 'undefined' ? Math.random() * .2 - .1 :
|
|
weight;
|
|
this.gain = 1;
|
|
this.gater = null;
|
|
}
|
|
|
|
|
|
// squashing functions
|
|
Neuron.squash = {};
|
|
|
|
// eq. 5 & 5'
|
|
Neuron.squash.LOGISTIC = function(x, derivate) {
|
|
if (!derivate)
|
|
return 1 / (1 + Math.exp(-x));
|
|
var fx = Neuron.squash.LOGISTIC(x);
|
|
return fx * (1 - fx);
|
|
};
|
|
Neuron.squash.TANH = function(x, derivate) {
|
|
if (derivate)
|
|
return 1 - Math.pow(Neuron.squash.TANH(x), 2);
|
|
var eP = Math.exp(x);
|
|
var eN = 1 / eP;
|
|
return (eP - eN) / (eP + eN);
|
|
};
|
|
Neuron.squash.IDENTITY = function(x, derivate) {
|
|
return derivate ? 1 : x;
|
|
};
|
|
Neuron.squash.HLIM = function(x, derivate) {
|
|
return derivate ? 1 : x > 0 ? 1 : 0;
|
|
};
|
|
Neuron.squash.RELU = function(x, derivate) {
|
|
if (derivate)
|
|
return x > 0 ? 1 : 0;
|
|
return x > 0 ? x : 0;
|
|
};
|
|
|
|
// unique ID's
|
|
(function() {
|
|
var neurons = 0;
|
|
var connections = 0;
|
|
Neuron.uid = function() {
|
|
return neurons++;
|
|
}
|
|
Neuron.connection.uid = function() {
|
|
return connections++;
|
|
}
|
|
Neuron.quantity = function() {
|
|
return {
|
|
neurons: neurons,
|
|
connections: connections
|
|
}
|
|
}
|
|
})();
|
|
|
|
/* WEBPACK VAR INJECTION */}.call(exports, __webpack_require__(2)(module)))
|
|
|
|
/***/ },
|
|
/* 2 */
|
|
/***/ function(module, exports) {
|
|
|
|
module.exports = function(module) {
|
|
if(!module.webpackPolyfill) {
|
|
module.deprecate = function() {};
|
|
module.paths = [];
|
|
// module.parent = undefined by default
|
|
module.children = [];
|
|
module.webpackPolyfill = 1;
|
|
}
|
|
return module;
|
|
}
|
|
|
|
|
|
/***/ },
|
|
/* 3 */
|
|
/***/ function(module, exports, __webpack_require__) {
|
|
|
|
/* WEBPACK VAR INJECTION */(function(module) {// export
|
|
if (module) module.exports = Layer;
|
|
|
|
// import
|
|
var Neuron = __webpack_require__(1)
|
|
, Network = __webpack_require__(4)
|
|
|
|
/*******************************************************************************************
|
|
LAYER
|
|
*******************************************************************************************/
|
|
|
|
function Layer(size, label) {
|
|
this.size = size | 0;
|
|
this.list = [];
|
|
this.label = label || null;
|
|
this.connectedTo = [];
|
|
|
|
while (size--) {
|
|
var neuron = new Neuron();
|
|
this.list.push(neuron);
|
|
}
|
|
}
|
|
|
|
Layer.prototype = {
|
|
|
|
// activates all the neurons in the layer
|
|
activate: function(input) {
|
|
|
|
var activations = [];
|
|
|
|
if (typeof input != 'undefined') {
|
|
if (input.length != this.size)
|
|
throw new Error("INPUT size and LAYER size must be the same to activate!");
|
|
|
|
for (var id in this.list) {
|
|
var neuron = this.list[id];
|
|
var activation = neuron.activate(input[id]);
|
|
activations.push(activation);
|
|
}
|
|
} else {
|
|
for (var id in this.list) {
|
|
var neuron = this.list[id];
|
|
var activation = neuron.activate();
|
|
activations.push(activation);
|
|
}
|
|
}
|
|
return activations;
|
|
},
|
|
|
|
// propagates the error on all the neurons of the layer
|
|
propagate: function(rate, target) {
|
|
|
|
if (typeof target != 'undefined') {
|
|
if (target.length != this.size)
|
|
throw new Error("TARGET size and LAYER size must be the same to propagate!");
|
|
|
|
for (var id = this.list.length - 1; id >= 0; id--) {
|
|
var neuron = this.list[id];
|
|
neuron.propagate(rate, target[id]);
|
|
}
|
|
} else {
|
|
for (var id = this.list.length - 1; id >= 0; id--) {
|
|
var neuron = this.list[id];
|
|
neuron.propagate(rate);
|
|
}
|
|
}
|
|
},
|
|
|
|
// projects a connection from this layer to another one
|
|
project: function(layer, type, weights) {
|
|
|
|
if (layer instanceof Network)
|
|
layer = layer.layers.input;
|
|
|
|
if (layer instanceof Layer) {
|
|
if (!this.connected(layer))
|
|
return new Layer.connection(this, layer, type, weights);
|
|
} else
|
|
throw new Error("Invalid argument, you can only project connections to LAYERS and NETWORKS!");
|
|
|
|
|
|
},
|
|
|
|
// gates a connection betwenn two layers
|
|
gate: function(connection, type) {
|
|
|
|
if (type == Layer.gateType.INPUT) {
|
|
if (connection.to.size != this.size)
|
|
throw new Error("GATER layer and CONNECTION.TO layer must be the same size in order to gate!");
|
|
|
|
for (var id in connection.to.list) {
|
|
var neuron = connection.to.list[id];
|
|
var gater = this.list[id];
|
|
for (var input in neuron.connections.inputs) {
|
|
var gated = neuron.connections.inputs[input];
|
|
if (gated.ID in connection.connections)
|
|
gater.gate(gated);
|
|
}
|
|
}
|
|
} else if (type == Layer.gateType.OUTPUT) {
|
|
if (connection.from.size != this.size)
|
|
throw new Error("GATER layer and CONNECTION.FROM layer must be the same size in order to gate!");
|
|
|
|
for (var id in connection.from.list) {
|
|
var neuron = connection.from.list[id];
|
|
var gater = this.list[id];
|
|
for (var projected in neuron.connections.projected) {
|
|
var gated = neuron.connections.projected[projected];
|
|
if (gated.ID in connection.connections)
|
|
gater.gate(gated);
|
|
}
|
|
}
|
|
} else if (type == Layer.gateType.ONE_TO_ONE) {
|
|
if (connection.size != this.size)
|
|
throw new Error("The number of GATER UNITS must be the same as the number of CONNECTIONS to gate!");
|
|
|
|
for (var id in connection.list) {
|
|
var gater = this.list[id];
|
|
var gated = connection.list[id];
|
|
gater.gate(gated);
|
|
}
|
|
}
|
|
connection.gatedfrom.push({layer: this, type: type});
|
|
},
|
|
|
|
// true or false whether the whole layer is self-connected or not
|
|
selfconnected: function() {
|
|
|
|
for (var id in this.list) {
|
|
var neuron = this.list[id];
|
|
if (!neuron.selfconnected())
|
|
return false;
|
|
}
|
|
return true;
|
|
},
|
|
|
|
// true of false whether the layer is connected to another layer (parameter) or not
|
|
connected: function(layer) {
|
|
// Check if ALL to ALL connection
|
|
var connections = 0;
|
|
for (var here in this.list) {
|
|
for (var there in layer.list) {
|
|
var from = this.list[here];
|
|
var to = layer.list[there];
|
|
var connected = from.connected(to);
|
|
if (connected.type == 'projected')
|
|
connections++;
|
|
}
|
|
}
|
|
if (connections == this.size * layer.size)
|
|
return Layer.connectionType.ALL_TO_ALL;
|
|
|
|
// Check if ONE to ONE connection
|
|
connections = 0;
|
|
for (var neuron in this.list) {
|
|
var from = this.list[neuron];
|
|
var to = layer.list[neuron];
|
|
var connected = from.connected(to);
|
|
if (connected.type == 'projected')
|
|
connections++;
|
|
}
|
|
if (connections == this.size)
|
|
return Layer.connectionType.ONE_TO_ONE;
|
|
},
|
|
|
|
// clears all the neuorns in the layer
|
|
clear: function() {
|
|
for (var id in this.list) {
|
|
var neuron = this.list[id];
|
|
neuron.clear();
|
|
}
|
|
},
|
|
|
|
// resets all the neurons in the layer
|
|
reset: function() {
|
|
for (var id in this.list) {
|
|
var neuron = this.list[id];
|
|
neuron.reset();
|
|
}
|
|
},
|
|
|
|
// returns all the neurons in the layer (array)
|
|
neurons: function() {
|
|
return this.list;
|
|
},
|
|
|
|
// adds a neuron to the layer
|
|
add: function(neuron) {
|
|
this.neurons[neuron.ID] = neuron || new Neuron();
|
|
this.list.push(neuron);
|
|
this.size++;
|
|
},
|
|
|
|
set: function(options) {
|
|
options = options || {};
|
|
|
|
for (var i in this.list) {
|
|
var neuron = this.list[i];
|
|
if (options.label)
|
|
neuron.label = options.label + '_' + neuron.ID;
|
|
if (options.squash)
|
|
neuron.squash = options.squash;
|
|
if (options.bias)
|
|
neuron.bias = options.bias;
|
|
}
|
|
return this;
|
|
}
|
|
}
|
|
|
|
// represents a connection from one layer to another, and keeps track of its weight and gain
|
|
Layer.connection = function LayerConnection(fromLayer, toLayer, type, weights) {
|
|
this.ID = Layer.connection.uid();
|
|
this.from = fromLayer;
|
|
this.to = toLayer;
|
|
this.selfconnection = toLayer == fromLayer;
|
|
this.type = type;
|
|
this.connections = {};
|
|
this.list = [];
|
|
this.size = 0;
|
|
this.gatedfrom = [];
|
|
|
|
if (typeof this.type == 'undefined')
|
|
{
|
|
if (fromLayer == toLayer)
|
|
this.type = Layer.connectionType.ONE_TO_ONE;
|
|
else
|
|
this.type = Layer.connectionType.ALL_TO_ALL;
|
|
}
|
|
|
|
if (this.type == Layer.connectionType.ALL_TO_ALL ||
|
|
this.type == Layer.connectionType.ALL_TO_ELSE) {
|
|
for (var here in this.from.list) {
|
|
for (var there in this.to.list) {
|
|
var from = this.from.list[here];
|
|
var to = this.to.list[there];
|
|
if(this.type == Layer.connectionType.ALL_TO_ELSE && from == to)
|
|
continue;
|
|
var connection = from.project(to, weights);
|
|
|
|
this.connections[connection.ID] = connection;
|
|
this.size = this.list.push(connection);
|
|
}
|
|
}
|
|
} else if (this.type == Layer.connectionType.ONE_TO_ONE) {
|
|
|
|
for (var neuron in this.from.list) {
|
|
var from = this.from.list[neuron];
|
|
var to = this.to.list[neuron];
|
|
var connection = from.project(to, weights);
|
|
|
|
this.connections[connection.ID] = connection;
|
|
this.size = this.list.push(connection);
|
|
}
|
|
}
|
|
|
|
fromLayer.connectedTo.push(this);
|
|
}
|
|
|
|
// types of connections
|
|
Layer.connectionType = {};
|
|
Layer.connectionType.ALL_TO_ALL = "ALL TO ALL";
|
|
Layer.connectionType.ONE_TO_ONE = "ONE TO ONE";
|
|
Layer.connectionType.ALL_TO_ELSE = "ALL TO ELSE";
|
|
|
|
// types of gates
|
|
Layer.gateType = {};
|
|
Layer.gateType.INPUT = "INPUT";
|
|
Layer.gateType.OUTPUT = "OUTPUT";
|
|
Layer.gateType.ONE_TO_ONE = "ONE TO ONE";
|
|
|
|
(function() {
|
|
var connections = 0;
|
|
Layer.connection.uid = function() {
|
|
return connections++;
|
|
}
|
|
})();
|
|
|
|
/* WEBPACK VAR INJECTION */}.call(exports, __webpack_require__(2)(module)))
|
|
|
|
/***/ },
|
|
/* 4 */
|
|
/***/ function(module, exports, __webpack_require__) {
|
|
|
|
/* WEBPACK VAR INJECTION */(function(module) {// export
|
|
if (module) module.exports = Network;
|
|
|
|
// import
|
|
var Neuron = __webpack_require__(1)
|
|
, Layer = __webpack_require__(3)
|
|
, Trainer = __webpack_require__(5)
|
|
|
|
/*******************************************************************************************
|
|
NETWORK
|
|
*******************************************************************************************/
|
|
|
|
function Network(layers) {
|
|
if (typeof layers != 'undefined') {
|
|
this.layers = layers || {
|
|
input: null,
|
|
hidden: {},
|
|
output: null
|
|
};
|
|
this.optimized = null;
|
|
}
|
|
}
|
|
Network.prototype = {
|
|
|
|
// feed-forward activation of all the layers to produce an ouput
|
|
activate: function(input) {
|
|
|
|
if (this.optimized === false)
|
|
{
|
|
this.layers.input.activate(input);
|
|
for (var layer in this.layers.hidden)
|
|
this.layers.hidden[layer].activate();
|
|
return this.layers.output.activate();
|
|
}
|
|
else
|
|
{
|
|
if (this.optimized == null)
|
|
this.optimize();
|
|
return this.optimized.activate(input);
|
|
}
|
|
},
|
|
|
|
// back-propagate the error thru the network
|
|
propagate: function(rate, target) {
|
|
|
|
if (this.optimized === false)
|
|
{
|
|
this.layers.output.propagate(rate, target);
|
|
var reverse = [];
|
|
for (var layer in this.layers.hidden)
|
|
reverse.push(this.layers.hidden[layer]);
|
|
reverse.reverse();
|
|
for (var layer in reverse)
|
|
reverse[layer].propagate(rate);
|
|
}
|
|
else
|
|
{
|
|
if (this.optimized == null)
|
|
this.optimize();
|
|
this.optimized.propagate(rate, target);
|
|
}
|
|
},
|
|
|
|
// project a connection to another unit (either a network or a layer)
|
|
project: function(unit, type, weights) {
|
|
|
|
if (this.optimized)
|
|
this.optimized.reset();
|
|
|
|
if (unit instanceof Network)
|
|
return this.layers.output.project(unit.layers.input, type, weights);
|
|
|
|
if (unit instanceof Layer)
|
|
return this.layers.output.project(unit, type, weights);
|
|
|
|
throw new Error("Invalid argument, you can only project connections to LAYERS and NETWORKS!");
|
|
},
|
|
|
|
// let this network gate a connection
|
|
gate: function(connection, type) {
|
|
if (this.optimized)
|
|
this.optimized.reset();
|
|
this.layers.output.gate(connection, type);
|
|
},
|
|
|
|
// clear all elegibility traces and extended elegibility traces (the network forgets its context, but not what was trained)
|
|
clear: function() {
|
|
|
|
this.restore();
|
|
|
|
var inputLayer = this.layers.input,
|
|
outputLayer = this.layers.output;
|
|
|
|
inputLayer.clear();
|
|
for (var layer in this.layers.hidden) {
|
|
var hiddenLayer = this.layers.hidden[layer];
|
|
hiddenLayer.clear();
|
|
}
|
|
outputLayer.clear();
|
|
|
|
if (this.optimized)
|
|
this.optimized.reset();
|
|
},
|
|
|
|
// reset all weights and clear all traces (ends up like a new network)
|
|
reset: function() {
|
|
|
|
this.restore();
|
|
|
|
var inputLayer = this.layers.input,
|
|
outputLayer = this.layers.output;
|
|
|
|
inputLayer.reset();
|
|
for (var layer in this.layers.hidden) {
|
|
var hiddenLayer = this.layers.hidden[layer];
|
|
hiddenLayer.reset();
|
|
}
|
|
outputLayer.reset();
|
|
|
|
if (this.optimized)
|
|
this.optimized.reset();
|
|
},
|
|
|
|
// hardcodes the behaviour of the whole network into a single optimized function
|
|
optimize: function() {
|
|
|
|
var that = this;
|
|
var optimized = {};
|
|
var neurons = this.neurons();
|
|
|
|
for (var i in neurons) {
|
|
var neuron = neurons[i].neuron;
|
|
var layer = neurons[i].layer;
|
|
while (neuron.neuron)
|
|
neuron = neuron.neuron;
|
|
optimized = neuron.optimize(optimized, layer);
|
|
}
|
|
for (var i in optimized.propagation_sentences)
|
|
optimized.propagation_sentences[i].reverse();
|
|
optimized.propagation_sentences.reverse();
|
|
|
|
var hardcode = "";
|
|
hardcode += "var F = Float64Array ? new Float64Array(" + optimized.memory +
|
|
") : []; ";
|
|
for (var i in optimized.variables)
|
|
hardcode += "F[" + optimized.variables[i].id + "] = " + (optimized.variables[
|
|
i].value || 0) + "; ";
|
|
hardcode += "var activate = function(input){\n";
|
|
for (var i in optimized.inputs)
|
|
hardcode += "F[" + optimized.inputs[i] + "] = input[" + i + "]; ";
|
|
for (var currentLayer in optimized.activation_sentences) {
|
|
if (optimized.activation_sentences[currentLayer].length > 0) {
|
|
for (var currentNeuron in optimized.activation_sentences[currentLayer]) {
|
|
hardcode += optimized.activation_sentences[currentLayer][currentNeuron].join(" ");
|
|
hardcode += optimized.trace_sentences[currentLayer][currentNeuron].join(" ");
|
|
}
|
|
}
|
|
}
|
|
hardcode += " var output = []; "
|
|
for (var i in optimized.outputs)
|
|
hardcode += "output[" + i + "] = F[" + optimized.outputs[i] + "]; ";
|
|
hardcode += "return output; }; "
|
|
hardcode += "var propagate = function(rate, target){\n";
|
|
hardcode += "F[" + optimized.variables.rate.id + "] = rate; ";
|
|
for (var i in optimized.targets)
|
|
hardcode += "F[" + optimized.targets[i] + "] = target[" + i + "]; ";
|
|
for (var currentLayer in optimized.propagation_sentences)
|
|
for (var currentNeuron in optimized.propagation_sentences[currentLayer])
|
|
hardcode += optimized.propagation_sentences[currentLayer][currentNeuron].join(" ") + " ";
|
|
hardcode += " };\n";
|
|
hardcode +=
|
|
"var ownership = function(memoryBuffer){\nF = memoryBuffer;\nthis.memory = F;\n};\n";
|
|
hardcode +=
|
|
"return {\nmemory: F,\nactivate: activate,\npropagate: propagate,\nownership: ownership\n};";
|
|
hardcode = hardcode.split(";").join(";\n");
|
|
|
|
var constructor = new Function(hardcode);
|
|
|
|
var network = constructor();
|
|
network.data = {
|
|
variables: optimized.variables,
|
|
activate: optimized.activation_sentences,
|
|
propagate: optimized.propagation_sentences,
|
|
trace: optimized.trace_sentences,
|
|
inputs: optimized.inputs,
|
|
outputs: optimized.outputs,
|
|
check_activation: this.activate,
|
|
check_propagation: this.propagate
|
|
}
|
|
|
|
network.reset = function() {
|
|
if (that.optimized) {
|
|
that.optimized = null;
|
|
that.activate = network.data.check_activation;
|
|
that.propagate = network.data.check_propagation;
|
|
}
|
|
}
|
|
|
|
this.optimized = network;
|
|
this.activate = network.activate;
|
|
this.propagate = network.propagate;
|
|
},
|
|
|
|
// restores all the values from the optimized network the their respective objects in order to manipulate the network
|
|
restore: function() {
|
|
if (!this.optimized)
|
|
return;
|
|
|
|
var optimized = this.optimized;
|
|
|
|
var getValue = function() {
|
|
var args = Array.prototype.slice.call(arguments);
|
|
|
|
var unit = args.shift();
|
|
var prop = args.pop();
|
|
|
|
var id = prop + '_';
|
|
for (var property in args)
|
|
id += args[property] + '_';
|
|
id += unit.ID;
|
|
|
|
var memory = optimized.memory;
|
|
var variables = optimized.data.variables;
|
|
|
|
if (id in variables)
|
|
return memory[variables[id].id];
|
|
return 0;
|
|
}
|
|
|
|
var list = this.neurons();
|
|
|
|
// link id's to positions in the array
|
|
var ids = {};
|
|
for (var i in list) {
|
|
var neuron = list[i].neuron;
|
|
while (neuron.neuron)
|
|
neuron = neuron.neuron;
|
|
|
|
neuron.state = getValue(neuron, 'state');
|
|
neuron.old = getValue(neuron, 'old');
|
|
neuron.activation = getValue(neuron, 'activation');
|
|
neuron.bias = getValue(neuron, 'bias');
|
|
|
|
for (var input in neuron.trace.elegibility)
|
|
neuron.trace.elegibility[input] = getValue(neuron, 'trace',
|
|
'elegibility', input);
|
|
|
|
for (var gated in neuron.trace.extended)
|
|
for (var input in neuron.trace.extended[gated])
|
|
neuron.trace.extended[gated][input] = getValue(neuron, 'trace',
|
|
'extended', gated, input);
|
|
}
|
|
|
|
// get connections
|
|
for (var i in list) {
|
|
var neuron = list[i].neuron;
|
|
while (neuron.neuron)
|
|
neuron = neuron.neuron;
|
|
|
|
for (var j in neuron.connections.projected) {
|
|
var connection = neuron.connections.projected[j];
|
|
connection.weight = getValue(connection, 'weight');
|
|
connection.gain = getValue(connection, 'gain');
|
|
}
|
|
}
|
|
},
|
|
|
|
// returns all the neurons in the network
|
|
neurons: function() {
|
|
|
|
var neurons = [];
|
|
|
|
var inputLayer = this.layers.input.neurons(),
|
|
outputLayer = this.layers.output.neurons();
|
|
|
|
for (var neuron in inputLayer)
|
|
neurons.push({
|
|
neuron: inputLayer[neuron],
|
|
layer: 'input'
|
|
});
|
|
|
|
for (var layer in this.layers.hidden) {
|
|
var hiddenLayer = this.layers.hidden[layer].neurons();
|
|
for (var neuron in hiddenLayer)
|
|
neurons.push({
|
|
neuron: hiddenLayer[neuron],
|
|
layer: layer
|
|
});
|
|
}
|
|
for (var neuron in outputLayer)
|
|
neurons.push({
|
|
neuron: outputLayer[neuron],
|
|
layer: 'output'
|
|
});
|
|
|
|
return neurons;
|
|
},
|
|
|
|
// returns number of inputs of the network
|
|
inputs: function() {
|
|
return this.layers.input.size;
|
|
},
|
|
|
|
// returns number of outputs of hte network
|
|
outputs: function() {
|
|
return this.layers.output.size;
|
|
},
|
|
|
|
// sets the layers of the network
|
|
set: function(layers) {
|
|
|
|
this.layers = layers;
|
|
if (this.optimized)
|
|
this.optimized.reset();
|
|
},
|
|
|
|
setOptimize: function(bool){
|
|
this.restore();
|
|
if (this.optimized)
|
|
this.optimized.reset();
|
|
this.optimized = bool? null : false;
|
|
},
|
|
|
|
// returns a json that represents all the neurons and connections of the network
|
|
toJSON: function(ignoreTraces) {
|
|
|
|
this.restore();
|
|
|
|
var list = this.neurons();
|
|
var neurons = [];
|
|
var connections = [];
|
|
|
|
// link id's to positions in the array
|
|
var ids = {};
|
|
for (var i in list) {
|
|
var neuron = list[i].neuron;
|
|
while (neuron.neuron)
|
|
neuron = neuron.neuron;
|
|
ids[neuron.ID] = i;
|
|
|
|
var copy = {
|
|
trace: {
|
|
elegibility: {},
|
|
extended: {}
|
|
},
|
|
state: neuron.state,
|
|
old: neuron.old,
|
|
activation: neuron.activation,
|
|
bias: neuron.bias,
|
|
layer: list[i].layer
|
|
};
|
|
|
|
copy.squash = neuron.squash == Neuron.squash.LOGISTIC ? "LOGISTIC" :
|
|
neuron.squash == Neuron.squash.TANH ? "TANH" :
|
|
neuron.squash == Neuron.squash.IDENTITY ? "IDENTITY" :
|
|
neuron.squash == Neuron.squash.HLIM ? "HLIM" :
|
|
null;
|
|
|
|
neurons.push(copy);
|
|
}
|
|
|
|
// get connections
|
|
for (var i in list) {
|
|
var neuron = list[i].neuron;
|
|
while (neuron.neuron)
|
|
neuron = neuron.neuron;
|
|
|
|
for (var j in neuron.connections.projected) {
|
|
var connection = neuron.connections.projected[j];
|
|
connections.push({
|
|
from: ids[connection.from.ID],
|
|
to: ids[connection.to.ID],
|
|
weight: connection.weight,
|
|
gater: connection.gater ? ids[connection.gater.ID] : null,
|
|
});
|
|
}
|
|
if (neuron.selfconnected())
|
|
connections.push({
|
|
from: ids[neuron.ID],
|
|
to: ids[neuron.ID],
|
|
weight: neuron.selfconnection.weight,
|
|
gater: neuron.selfconnection.gater ? ids[neuron.selfconnection.gater.ID] : null,
|
|
});
|
|
}
|
|
|
|
return {
|
|
neurons: neurons,
|
|
connections: connections
|
|
}
|
|
},
|
|
|
|
// export the topology into dot language which can be visualized as graphs using dot
|
|
/* example: ... console.log(net.toDotLang());
|
|
$ node example.js > example.dot
|
|
$ dot example.dot -Tpng > out.png
|
|
*/
|
|
toDot: function(edgeConnection) {
|
|
if (! typeof edgeConnection)
|
|
edgeConnection = false;
|
|
var code = "digraph nn {\n rankdir = BT\n";
|
|
var layers = [this.layers.input].concat(this.layers.hidden, this.layers.output);
|
|
for (var layer in layers) {
|
|
for (var to in layers[layer].connectedTo) { // projections
|
|
var connection = layers[layer].connectedTo[to];
|
|
var layerTo = connection.to;
|
|
var size = connection.size;
|
|
var layerID = layers.indexOf(layers[layer]);
|
|
var layerToID = layers.indexOf(layerTo);
|
|
/* http://stackoverflow.com/questions/26845540/connect-edges-with-graph-dot
|
|
* DOT does not support edge-to-edge connections
|
|
* This workaround produces somewhat weird graphs ...
|
|
*/
|
|
if ( edgeConnection) {
|
|
if (connection.gatedfrom.length) {
|
|
var fakeNode = "fake" + layerID + "_" + layerToID;
|
|
code += " " + fakeNode +
|
|
" [label = \"\", shape = point, width = 0.01, height = 0.01]\n";
|
|
code += " " + layerID + " -> " + fakeNode + " [label = " + size + ", arrowhead = none]\n";
|
|
code += " " + fakeNode + " -> " + layerToID + "\n";
|
|
} else
|
|
code += " " + layerID + " -> " + layerToID + " [label = " + size + "]\n";
|
|
for (var from in connection.gatedfrom) { // gatings
|
|
var layerfrom = connection.gatedfrom[from].layer;
|
|
var layerfromID = layers.indexOf(layerfrom);
|
|
code += " " + layerfromID + " -> " + fakeNode + " [color = blue]\n";
|
|
}
|
|
} else {
|
|
code += " " + layerID + " -> " + layerToID + " [label = " + size + "]\n";
|
|
for (var from in connection.gatedfrom) { // gatings
|
|
var layerfrom = connection.gatedfrom[from].layer;
|
|
var layerfromID = layers.indexOf(layerfrom);
|
|
code += " " + layerfromID + " -> " + layerToID + " [color = blue]\n";
|
|
}
|
|
}
|
|
}
|
|
}
|
|
code += "}\n";
|
|
return {
|
|
code: code,
|
|
link: "https://chart.googleapis.com/chart?chl=" + escape(code.replace("/ /g", "+")) + "&cht=gv"
|
|
}
|
|
},
|
|
|
|
// returns a function that works as the activation of the network and can be used without depending on the library
|
|
standalone: function() {
|
|
if (!this.optimized)
|
|
this.optimize();
|
|
|
|
var data = this.optimized.data;
|
|
|
|
// build activation function
|
|
var activation = "function (input) {\n";
|
|
|
|
// build inputs
|
|
for (var i in data.inputs)
|
|
activation += "F[" + data.inputs[i] + "] = input[" + i + "];\n";
|
|
|
|
// build network activation
|
|
for (var neuron in data.activate) { // shouldn't this be layer?
|
|
for (var sentence in data.activate[neuron])
|
|
activation += data.activate[neuron][sentence].join('') + "\n";
|
|
}
|
|
|
|
// build outputs
|
|
activation += "var output = [];\n";
|
|
for (var i in data.outputs)
|
|
activation += "output[" + i + "] = F[" + data.outputs[i] + "];\n";
|
|
activation += "return output;\n}";
|
|
|
|
// reference all the positions in memory
|
|
var memory = activation.match(/F\[(\d+)\]/g);
|
|
var dimension = 0;
|
|
var ids = {};
|
|
for (var address in memory) {
|
|
var tmp = memory[address].match(/\d+/)[0];
|
|
if (!(tmp in ids)) {
|
|
ids[tmp] = dimension++;
|
|
}
|
|
}
|
|
var hardcode = "F = {\n";
|
|
for (var i in ids)
|
|
hardcode += ids[i] + ": " + this.optimized.memory[i] + ",\n";
|
|
hardcode = hardcode.substring(0, hardcode.length - 2) + "\n};\n";
|
|
hardcode = "var run = " + activation.replace(/F\[(\d+)]/g, function(
|
|
index) {
|
|
return 'F[' + ids[index.match(/\d+/)[0]] + ']'
|
|
}).replace("{\n", "{\n" + hardcode + "") + ";\n";
|
|
hardcode += "return run";
|
|
|
|
// return standalone function
|
|
return new Function(hardcode)();
|
|
},
|
|
|
|
|
|
// Return a HTML5 WebWorker specialized on training the network stored in `memory`.
|
|
// Train based on the given dataSet and options.
|
|
// The worker returns the updated `memory` when done.
|
|
worker: function(memory, set, options) {
|
|
|
|
// Copy the options and set defaults (options might be different for each worker)
|
|
var workerOptions = {};
|
|
if(options) workerOptions = options;
|
|
workerOptions.rate = options.rate || .2;
|
|
workerOptions.iterations = options.iterations || 100000;
|
|
workerOptions.error = options.error || .005;
|
|
workerOptions.cost = options.cost || null;
|
|
workerOptions.crossValidate = options.crossValidate || null;
|
|
|
|
// Cost function might be different for each worker
|
|
costFunction = "var cost = " + (options && options.cost || this.cost || Trainer.cost.MSE) + ";\n";
|
|
var workerFunction = Network.getWorkerSharedFunctions();
|
|
workerFunction = workerFunction.replace(/var cost = options && options\.cost \|\| this\.cost \|\| Trainer\.cost\.MSE;/g, costFunction);
|
|
|
|
// Set what we do when training is finished
|
|
workerFunction = workerFunction.replace('return results;',
|
|
'postMessage({action: "done", message: results, memoryBuffer: F}, [F.buffer]);');
|
|
|
|
// Replace log with postmessage
|
|
workerFunction = workerFunction.replace("console.log('iterations', iterations, 'error', error, 'rate', currentRate)",
|
|
"postMessage({action: 'log', message: {\n" +
|
|
"iterations: iterations,\n" +
|
|
"error: error,\n" +
|
|
"rate: currentRate\n" +
|
|
"}\n" +
|
|
"})");
|
|
|
|
// Replace schedule with postmessage
|
|
workerFunction = workerFunction.replace("abort = this.schedule.do({ error: error, iterations: iterations, rate: currentRate })",
|
|
"postMessage({action: 'schedule', message: {\n" +
|
|
"iterations: iterations,\n" +
|
|
"error: error,\n" +
|
|
"rate: currentRate\n" +
|
|
"}\n" +
|
|
"})");
|
|
|
|
if (!this.optimized)
|
|
this.optimize();
|
|
|
|
var hardcode = "var inputs = " + this.optimized.data.inputs.length + ";\n";
|
|
hardcode += "var outputs = " + this.optimized.data.outputs.length + ";\n";
|
|
hardcode += "var F = new Float64Array([" + this.optimized.memory.toString() + "]);\n";
|
|
hardcode += "var activate = " + this.optimized.activate.toString() + ";\n";
|
|
hardcode += "var propagate = " + this.optimized.propagate.toString() + ";\n";
|
|
hardcode +=
|
|
"onmessage = function(e) {\n" +
|
|
"if (e.data.action == 'startTraining') {\n" +
|
|
"train(" + JSON.stringify(set) + "," + JSON.stringify(workerOptions) + ");\n" +
|
|
"}\n" +
|
|
"}";
|
|
|
|
var workerSourceCode = workerFunction + '\n' + hardcode;
|
|
var blob = new Blob([workerSourceCode]);
|
|
var blobURL = window.URL.createObjectURL(blob);
|
|
|
|
return new Worker(blobURL);
|
|
},
|
|
|
|
// returns a copy of the network
|
|
clone: function() {
|
|
return Network.fromJSON(this.toJSON());
|
|
}
|
|
};
|
|
|
|
/**
|
|
* Creates a static String to store the source code of the functions
|
|
* that are identical for all the workers (train, _trainSet, test)
|
|
*
|
|
* @return {String} Source code that can train a network inside a worker.
|
|
* @static
|
|
*/
|
|
Network.getWorkerSharedFunctions = function() {
|
|
// If we already computed the source code for the shared functions
|
|
if(typeof Network._SHARED_WORKER_FUNCTIONS !== 'undefined')
|
|
return Network._SHARED_WORKER_FUNCTIONS;
|
|
|
|
// Otherwise compute and return the source code
|
|
// We compute them by simply copying the source code of the train, _trainSet and test functions
|
|
// using the .toString() method
|
|
|
|
// Load and name the train function
|
|
var train_f = Trainer.prototype.train.toString();
|
|
train_f = train_f.replace('function (set', 'function train(set') + '\n';
|
|
|
|
// Load and name the _trainSet function
|
|
var _trainSet_f = Trainer.prototype._trainSet.toString().replace(/this.network./g, '');
|
|
_trainSet_f = _trainSet_f.replace('function (set', 'function _trainSet(set') + '\n';
|
|
_trainSet_f = _trainSet_f.replace('this.crossValidate', 'crossValidate');
|
|
_trainSet_f = _trainSet_f.replace('crossValidate = true', 'crossValidate = { }');
|
|
|
|
// Load and name the test function
|
|
var test_f = Trainer.prototype.test.toString().replace(/this.network./g, '');
|
|
test_f = test_f.replace('function (set', 'function test(set') + '\n';
|
|
|
|
return Network._SHARED_WORKER_FUNCTIONS = train_f + _trainSet_f + test_f;
|
|
};
|
|
|
|
// rebuild a network that has been stored in a json using the method toJSON()
|
|
Network.fromJSON = function(json) {
|
|
|
|
var neurons = [];
|
|
|
|
var layers = {
|
|
input: new Layer(),
|
|
hidden: [],
|
|
output: new Layer()
|
|
};
|
|
|
|
for (var i in json.neurons) {
|
|
var config = json.neurons[i];
|
|
|
|
var neuron = new Neuron();
|
|
neuron.trace.elegibility = {};
|
|
neuron.trace.extended = {};
|
|
neuron.state = config.state;
|
|
neuron.old = config.old;
|
|
neuron.activation = config.activation;
|
|
neuron.bias = config.bias;
|
|
neuron.squash = config.squash in Neuron.squash ? Neuron.squash[config.squash] : Neuron.squash.LOGISTIC;
|
|
neurons.push(neuron);
|
|
|
|
if (config.layer == 'input')
|
|
layers.input.add(neuron);
|
|
else if (config.layer == 'output')
|
|
layers.output.add(neuron);
|
|
else {
|
|
if (typeof layers.hidden[config.layer] == 'undefined')
|
|
layers.hidden[config.layer] = new Layer();
|
|
layers.hidden[config.layer].add(neuron);
|
|
}
|
|
}
|
|
|
|
for (var i in json.connections) {
|
|
var config = json.connections[i];
|
|
var from = neurons[config.from];
|
|
var to = neurons[config.to];
|
|
var weight = config.weight;
|
|
var gater = neurons[config.gater];
|
|
|
|
var connection = from.project(to, weight);
|
|
if (gater)
|
|
gater.gate(connection);
|
|
}
|
|
|
|
return new Network(layers);
|
|
};
|
|
|
|
/* WEBPACK VAR INJECTION */}.call(exports, __webpack_require__(2)(module)))
|
|
|
|
/***/ },
|
|
/* 5 */
|
|
/***/ function(module, exports, __webpack_require__) {
|
|
|
|
/* WEBPACK VAR INJECTION */(function(module) {// export
|
|
if (module) module.exports = Trainer;
|
|
|
|
/*******************************************************************************************
|
|
TRAINER
|
|
*******************************************************************************************/
|
|
|
|
function Trainer(network, options) {
|
|
options = options || {};
|
|
this.network = network;
|
|
this.rate = options.rate || .2;
|
|
this.iterations = options.iterations || 100000;
|
|
this.error = options.error || .005;
|
|
this.cost = options.cost || null;
|
|
this.crossValidate = options.crossValidate || null;
|
|
}
|
|
|
|
Trainer.prototype = {
|
|
|
|
// trains any given set to a network
|
|
train: function(set, options) {
|
|
|
|
var error = 1;
|
|
var iterations = bucketSize = 0;
|
|
var abort = false;
|
|
var currentRate;
|
|
var cost = options && options.cost || this.cost || Trainer.cost.MSE;
|
|
var crossValidate = false, testSet, trainSet;
|
|
|
|
var start = Date.now();
|
|
|
|
if (options) {
|
|
if (options.shuffle) {
|
|
//+ Jonas Raoni Soares Silva
|
|
//@ http://jsfromhell.com/array/shuffle [v1.0]
|
|
function shuffle(o) { //v1.0
|
|
for (var j, x, i = o.length; i; j = Math.floor(Math.random() * i), x = o[--i], o[i] = o[j], o[j] = x);
|
|
return o;
|
|
};
|
|
}
|
|
if (options.iterations)
|
|
this.iterations = options.iterations;
|
|
if (options.error)
|
|
this.error = options.error;
|
|
if (options.rate)
|
|
this.rate = options.rate;
|
|
if (options.cost)
|
|
this.cost = options.cost;
|
|
if (options.schedule)
|
|
this.schedule = options.schedule;
|
|
if (options.customLog){
|
|
// for backward compatibility with code that used customLog
|
|
console.log('Deprecated: use schedule instead of customLog')
|
|
this.schedule = options.customLog;
|
|
}
|
|
if (this.crossValidate || options.crossValidate) {
|
|
if(!this.crossValidate) this.crossValidate = {};
|
|
crossValidate = true;
|
|
if (options.crossValidate.testSize)
|
|
this.crossValidate.testSize = options.crossValidate.testSize;
|
|
if (options.crossValidate.testError)
|
|
this.crossValidate.testError = options.crossValidate.testError;
|
|
}
|
|
}
|
|
|
|
currentRate = this.rate;
|
|
if(Array.isArray(this.rate)) {
|
|
var bucketSize = Math.floor(this.iterations / this.rate.length);
|
|
}
|
|
|
|
if(crossValidate) {
|
|
var numTrain = Math.ceil((1 - this.crossValidate.testSize) * set.length);
|
|
trainSet = set.slice(0, numTrain);
|
|
testSet = set.slice(numTrain);
|
|
}
|
|
|
|
var lastError = 0;
|
|
while ((!abort && iterations < this.iterations && error > this.error)) {
|
|
if (crossValidate && error <= this.crossValidate.testError) {
|
|
break;
|
|
}
|
|
|
|
var currentSetSize = set.length;
|
|
error = 0;
|
|
iterations++;
|
|
|
|
if(bucketSize > 0) {
|
|
var currentBucket = Math.floor(iterations / bucketSize);
|
|
currentRate = this.rate[currentBucket] || currentRate;
|
|
}
|
|
|
|
if(typeof this.rate === 'function') {
|
|
currentRate = this.rate(iterations, lastError);
|
|
}
|
|
|
|
if (crossValidate) {
|
|
this._trainSet(trainSet, currentRate, cost);
|
|
error += this.test(testSet).error;
|
|
currentSetSize = 1;
|
|
} else {
|
|
error += this._trainSet(set, currentRate, cost);
|
|
currentSetSize = set.length;
|
|
}
|
|
|
|
// check error
|
|
error /= currentSetSize;
|
|
lastError = error;
|
|
|
|
if (options) {
|
|
if (this.schedule && this.schedule.every && iterations %
|
|
this.schedule.every == 0)
|
|
abort = this.schedule.do({ error: error, iterations: iterations, rate: currentRate });
|
|
else if (options.log && iterations % options.log == 0) {
|
|
console.log('iterations', iterations, 'error', error, 'rate', currentRate);
|
|
};
|
|
if (options.shuffle)
|
|
shuffle(set);
|
|
}
|
|
}
|
|
|
|
var results = {
|
|
error: error,
|
|
iterations: iterations,
|
|
time: Date.now() - start
|
|
};
|
|
|
|
return results;
|
|
},
|
|
|
|
// trains any given set to a network, using a WebWorker (only for the browser). Returns a Promise of the results.
|
|
trainAsync: function(set, options) {
|
|
var train = this.workerTrain.bind(this);
|
|
return new Promise(function(resolve, reject) {
|
|
try {
|
|
train(set, resolve, options, true)
|
|
} catch(e) {
|
|
reject(e)
|
|
}
|
|
})
|
|
},
|
|
|
|
// preforms one training epoch and returns the error (private function used in this.train)
|
|
_trainSet: function(set, currentRate, costFunction) {
|
|
var errorSum = 0;
|
|
for (var train in set) {
|
|
var input = set[train].input;
|
|
var target = set[train].output;
|
|
|
|
var output = this.network.activate(input);
|
|
this.network.propagate(currentRate, target);
|
|
|
|
errorSum += costFunction(target, output);
|
|
}
|
|
return errorSum;
|
|
},
|
|
|
|
// tests a set and returns the error and elapsed time
|
|
test: function(set, options) {
|
|
|
|
var error = 0;
|
|
var input, output, target;
|
|
var cost = options && options.cost || this.cost || Trainer.cost.MSE;
|
|
|
|
var start = Date.now();
|
|
|
|
for (var test in set) {
|
|
input = set[test].input;
|
|
target = set[test].output;
|
|
output = this.network.activate(input);
|
|
error += cost(target, output);
|
|
}
|
|
|
|
error /= set.length;
|
|
|
|
var results = {
|
|
error: error,
|
|
time: Date.now() - start
|
|
};
|
|
|
|
return results;
|
|
},
|
|
|
|
// trains any given set to a network using a WebWorker [deprecated: use trainAsync instead]
|
|
workerTrain: function(set, callback, options, suppressWarning) {
|
|
|
|
if (!suppressWarning) {
|
|
console.warn('Deprecated: do not use `workerTrain`, use `trainAsync` instead.')
|
|
}
|
|
var that = this;
|
|
|
|
if (!this.network.optimized)
|
|
this.network.optimize();
|
|
|
|
// Create a new worker
|
|
var worker = this.network.worker(this.network.optimized.memory, set, options);
|
|
|
|
// train the worker
|
|
worker.onmessage = function(e) {
|
|
switch(e.data.action) {
|
|
case 'done':
|
|
var iterations = e.data.message.iterations;
|
|
var error = e.data.message.error;
|
|
var time = e.data.message.time;
|
|
|
|
that.network.optimized.ownership(e.data.memoryBuffer);
|
|
|
|
// Done callback
|
|
callback({
|
|
error: error,
|
|
iterations: iterations,
|
|
time: time
|
|
});
|
|
|
|
// Delete the worker and all its associated memory
|
|
worker.terminate();
|
|
break;
|
|
|
|
case 'log':
|
|
console.log(e.data.message);
|
|
|
|
case 'schedule':
|
|
if (options && options.schedule && typeof options.schedule.do === 'function') {
|
|
var scheduled = options.schedule.do
|
|
scheduled(e.data.message)
|
|
}
|
|
break;
|
|
}
|
|
};
|
|
|
|
// Start the worker
|
|
worker.postMessage({action: 'startTraining'});
|
|
},
|
|
|
|
// trains an XOR to the network
|
|
XOR: function(options) {
|
|
|
|
if (this.network.inputs() != 2 || this.network.outputs() != 1)
|
|
throw new Error("Incompatible network (2 inputs, 1 output)");
|
|
|
|
var defaults = {
|
|
iterations: 100000,
|
|
log: false,
|
|
shuffle: true,
|
|
cost: Trainer.cost.MSE
|
|
};
|
|
|
|
if (options)
|
|
for (var i in options)
|
|
defaults[i] = options[i];
|
|
|
|
return this.train([{
|
|
input: [0, 0],
|
|
output: [0]
|
|
}, {
|
|
input: [1, 0],
|
|
output: [1]
|
|
}, {
|
|
input: [0, 1],
|
|
output: [1]
|
|
}, {
|
|
input: [1, 1],
|
|
output: [0]
|
|
}], defaults);
|
|
},
|
|
|
|
// trains the network to pass a Distracted Sequence Recall test
|
|
DSR: function(options) {
|
|
options = options || {};
|
|
|
|
var targets = options.targets || [2, 4, 7, 8];
|
|
var distractors = options.distractors || [3, 5, 6, 9];
|
|
var prompts = options.prompts || [0, 1];
|
|
var length = options.length || 24;
|
|
var criterion = options.success || 0.95;
|
|
var iterations = options.iterations || 100000;
|
|
var rate = options.rate || .1;
|
|
var log = options.log || 0;
|
|
var schedule = options.schedule || {};
|
|
var cost = options.cost || this.cost || Trainer.cost.CROSS_ENTROPY;
|
|
|
|
var trial, correct, i, j, success;
|
|
trial = correct = i = j = success = 0;
|
|
var error = 1,
|
|
symbols = targets.length + distractors.length + prompts.length;
|
|
|
|
var noRepeat = function(range, avoid) {
|
|
var number = Math.random() * range | 0;
|
|
var used = false;
|
|
for (var i in avoid)
|
|
if (number == avoid[i])
|
|
used = true;
|
|
return used ? noRepeat(range, avoid) : number;
|
|
};
|
|
|
|
var equal = function(prediction, output) {
|
|
for (var i in prediction)
|
|
if (Math.round(prediction[i]) != output[i])
|
|
return false;
|
|
return true;
|
|
};
|
|
|
|
var start = Date.now();
|
|
|
|
while (trial < iterations && (success < criterion || trial % 1000 != 0)) {
|
|
// generate sequence
|
|
var sequence = [],
|
|
sequenceLength = length - prompts.length;
|
|
for (i = 0; i < sequenceLength; i++) {
|
|
var any = Math.random() * distractors.length | 0;
|
|
sequence.push(distractors[any]);
|
|
}
|
|
var indexes = [],
|
|
positions = [];
|
|
for (i = 0; i < prompts.length; i++) {
|
|
indexes.push(Math.random() * targets.length | 0);
|
|
positions.push(noRepeat(sequenceLength, positions));
|
|
}
|
|
positions = positions.sort();
|
|
for (i = 0; i < prompts.length; i++) {
|
|
sequence[positions[i]] = targets[indexes[i]];
|
|
sequence.push(prompts[i]);
|
|
}
|
|
|
|
//train sequence
|
|
var distractorsCorrect;
|
|
var targetsCorrect = distractorsCorrect = 0;
|
|
error = 0;
|
|
for (i = 0; i < length; i++) {
|
|
// generate input from sequence
|
|
var input = [];
|
|
for (j = 0; j < symbols; j++)
|
|
input[j] = 0;
|
|
input[sequence[i]] = 1;
|
|
|
|
// generate target output
|
|
var output = [];
|
|
for (j = 0; j < targets.length; j++)
|
|
output[j] = 0;
|
|
|
|
if (i >= sequenceLength) {
|
|
var index = i - sequenceLength;
|
|
output[indexes[index]] = 1;
|
|
}
|
|
|
|
// check result
|
|
var prediction = this.network.activate(input);
|
|
|
|
if (equal(prediction, output))
|
|
if (i < sequenceLength)
|
|
distractorsCorrect++;
|
|
else
|
|
targetsCorrect++;
|
|
else {
|
|
this.network.propagate(rate, output);
|
|
}
|
|
|
|
error += cost(output, prediction);
|
|
|
|
if (distractorsCorrect + targetsCorrect == length)
|
|
correct++;
|
|
}
|
|
|
|
// calculate error
|
|
if (trial % 1000 == 0)
|
|
correct = 0;
|
|
trial++;
|
|
var divideError = trial % 1000;
|
|
divideError = divideError == 0 ? 1000 : divideError;
|
|
success = correct / divideError;
|
|
error /= length;
|
|
|
|
// log
|
|
if (log && trial % log == 0)
|
|
console.log("iterations:", trial, " success:", success, " correct:",
|
|
correct, " time:", Date.now() - start, " error:", error);
|
|
if (schedule.do && schedule.every && trial % schedule.every == 0)
|
|
schedule.do({
|
|
iterations: trial,
|
|
success: success,
|
|
error: error,
|
|
time: Date.now() - start,
|
|
correct: correct
|
|
});
|
|
}
|
|
|
|
return {
|
|
iterations: trial,
|
|
success: success,
|
|
error: error,
|
|
time: Date.now() - start
|
|
}
|
|
},
|
|
|
|
// train the network to learn an Embeded Reber Grammar
|
|
ERG: function(options) {
|
|
|
|
options = options || {};
|
|
var iterations = options.iterations || 150000;
|
|
var criterion = options.error || .05;
|
|
var rate = options.rate || .1;
|
|
var log = options.log || 500;
|
|
var cost = options.cost || this.cost || Trainer.cost.CROSS_ENTROPY;
|
|
|
|
// gramar node
|
|
var Node = function() {
|
|
this.paths = [];
|
|
};
|
|
Node.prototype = {
|
|
connect: function(node, value) {
|
|
this.paths.push({
|
|
node: node,
|
|
value: value
|
|
});
|
|
return this;
|
|
},
|
|
any: function() {
|
|
if (this.paths.length == 0)
|
|
return false;
|
|
var index = Math.random() * this.paths.length | 0;
|
|
return this.paths[index];
|
|
},
|
|
test: function(value) {
|
|
for (var i in this.paths)
|
|
if (this.paths[i].value == value)
|
|
return this.paths[i];
|
|
return false;
|
|
}
|
|
};
|
|
|
|
var reberGrammar = function() {
|
|
|
|
// build a reber grammar
|
|
var output = new Node();
|
|
var n1 = (new Node()).connect(output, "E");
|
|
var n2 = (new Node()).connect(n1, "S");
|
|
var n3 = (new Node()).connect(n1, "V").connect(n2, "P");
|
|
var n4 = (new Node()).connect(n2, "X");
|
|
n4.connect(n4, "S");
|
|
var n5 = (new Node()).connect(n3, "V");
|
|
n5.connect(n5, "T");
|
|
n2.connect(n5, "X");
|
|
var n6 = (new Node()).connect(n4, "T").connect(n5, "P");
|
|
var input = (new Node()).connect(n6, "B");
|
|
|
|
return {
|
|
input: input,
|
|
output: output
|
|
}
|
|
};
|
|
|
|
// build an embeded reber grammar
|
|
var embededReberGrammar = function() {
|
|
var reber1 = reberGrammar();
|
|
var reber2 = reberGrammar();
|
|
|
|
var output = new Node();
|
|
var n1 = (new Node).connect(output, "E");
|
|
reber1.output.connect(n1, "T");
|
|
reber2.output.connect(n1, "P");
|
|
var n2 = (new Node).connect(reber1.input, "P").connect(reber2.input,
|
|
"T");
|
|
var input = (new Node).connect(n2, "B");
|
|
|
|
return {
|
|
input: input,
|
|
output: output
|
|
}
|
|
|
|
};
|
|
|
|
// generate an ERG sequence
|
|
var generate = function() {
|
|
var node = embededReberGrammar().input;
|
|
var next = node.any();
|
|
var str = "";
|
|
while (next) {
|
|
str += next.value;
|
|
next = next.node.any();
|
|
}
|
|
return str;
|
|
};
|
|
|
|
// test if a string matches an embeded reber grammar
|
|
var test = function(str) {
|
|
var node = embededReberGrammar().input;
|
|
var i = 0;
|
|
var ch = str.charAt(i);
|
|
while (i < str.length) {
|
|
var next = node.test(ch);
|
|
if (!next)
|
|
return false;
|
|
node = next.node;
|
|
ch = str.charAt(++i);
|
|
}
|
|
return true;
|
|
};
|
|
|
|
// helper to check if the output and the target vectors match
|
|
var different = function(array1, array2) {
|
|
var max1 = 0;
|
|
var i1 = -1;
|
|
var max2 = 0;
|
|
var i2 = -1;
|
|
for (var i in array1) {
|
|
if (array1[i] > max1) {
|
|
max1 = array1[i];
|
|
i1 = i;
|
|
}
|
|
if (array2[i] > max2) {
|
|
max2 = array2[i];
|
|
i2 = i;
|
|
}
|
|
}
|
|
|
|
return i1 != i2;
|
|
};
|
|
|
|
var iteration = 0;
|
|
var error = 1;
|
|
var table = {
|
|
"B": 0,
|
|
"P": 1,
|
|
"T": 2,
|
|
"X": 3,
|
|
"S": 4,
|
|
"E": 5
|
|
};
|
|
|
|
var start = Date.now();
|
|
while (iteration < iterations && error > criterion) {
|
|
var i = 0;
|
|
error = 0;
|
|
|
|
// ERG sequence to learn
|
|
var sequence = generate();
|
|
|
|
// input
|
|
var read = sequence.charAt(i);
|
|
// target
|
|
var predict = sequence.charAt(i + 1);
|
|
|
|
// train
|
|
while (i < sequence.length - 1) {
|
|
var input = [];
|
|
var target = [];
|
|
for (var j = 0; j < 6; j++) {
|
|
input[j] = 0;
|
|
target[j] = 0;
|
|
}
|
|
input[table[read]] = 1;
|
|
target[table[predict]] = 1;
|
|
|
|
var output = this.network.activate(input);
|
|
|
|
if (different(output, target))
|
|
this.network.propagate(rate, target);
|
|
|
|
read = sequence.charAt(++i);
|
|
predict = sequence.charAt(i + 1);
|
|
|
|
error += cost(target, output);
|
|
}
|
|
error /= sequence.length;
|
|
iteration++;
|
|
if (iteration % log == 0) {
|
|
console.log("iterations:", iteration, " time:", Date.now() - start,
|
|
" error:", error);
|
|
}
|
|
}
|
|
|
|
return {
|
|
iterations: iteration,
|
|
error: error,
|
|
time: Date.now() - start,
|
|
test: test,
|
|
generate: generate
|
|
}
|
|
},
|
|
|
|
timingTask: function(options){
|
|
|
|
if (this.network.inputs() != 2 || this.network.outputs() != 1)
|
|
throw new Error("Invalid Network: must have 2 inputs and one output");
|
|
|
|
if (typeof options == 'undefined')
|
|
options = {};
|
|
|
|
// helper
|
|
function getSamples (trainingSize, testSize){
|
|
|
|
// sample size
|
|
var size = trainingSize + testSize;
|
|
|
|
// generate samples
|
|
var t = 0;
|
|
var set = [];
|
|
for (var i = 0; i < size; i++) {
|
|
set.push({ input: [0,0], output: [0] });
|
|
}
|
|
while(t < size - 20) {
|
|
var n = Math.round(Math.random() * 20);
|
|
set[t].input[0] = 1;
|
|
for (var j = t; j <= t + n; j++){
|
|
set[j].input[1] = n / 20;
|
|
set[j].output[0] = 0.5;
|
|
}
|
|
t += n;
|
|
n = Math.round(Math.random() * 20);
|
|
for (var k = t+1; k <= (t + n) && k < size; k++)
|
|
set[k].input[1] = set[t].input[1];
|
|
t += n;
|
|
}
|
|
|
|
// separate samples between train and test sets
|
|
var trainingSet = []; var testSet = [];
|
|
for (var l = 0; l < size; l++)
|
|
(l < trainingSize ? trainingSet : testSet).push(set[l]);
|
|
|
|
// return samples
|
|
return {
|
|
train: trainingSet,
|
|
test: testSet
|
|
}
|
|
}
|
|
|
|
var iterations = options.iterations || 200;
|
|
var error = options.error || .005;
|
|
var rate = options.rate || [.03, .02];
|
|
var log = options.log === false ? false : options.log || 10;
|
|
var cost = options.cost || this.cost || Trainer.cost.MSE;
|
|
var trainingSamples = options.trainSamples || 7000;
|
|
var testSamples = options.trainSamples || 1000;
|
|
|
|
// samples for training and testing
|
|
var samples = getSamples(trainingSamples, testSamples);
|
|
|
|
// train
|
|
var result = this.train(samples.train, {
|
|
rate: rate,
|
|
log: log,
|
|
iterations: iterations,
|
|
error: error,
|
|
cost: cost
|
|
});
|
|
|
|
return {
|
|
train: result,
|
|
test: this.test(samples.test)
|
|
}
|
|
}
|
|
};
|
|
|
|
// Built-in cost functions
|
|
Trainer.cost = {
|
|
// Eq. 9
|
|
CROSS_ENTROPY: function(target, output)
|
|
{
|
|
var crossentropy = 0;
|
|
for (var i in output)
|
|
crossentropy -= (target[i] * Math.log(output[i]+1e-15)) + ((1-target[i]) * Math.log((1+1e-15)-output[i])); // +1e-15 is a tiny push away to avoid Math.log(0)
|
|
return crossentropy;
|
|
},
|
|
MSE: function(target, output)
|
|
{
|
|
var mse = 0;
|
|
for (var i in output)
|
|
mse += Math.pow(target[i] - output[i], 2);
|
|
return mse / output.length;
|
|
},
|
|
BINARY: function(target, output){
|
|
var misses = 0;
|
|
for (var i in output)
|
|
misses += Math.round(target[i] * 2) != Math.round(output[i] * 2);
|
|
return misses;
|
|
}
|
|
}
|
|
|
|
/* WEBPACK VAR INJECTION */}.call(exports, __webpack_require__(2)(module)))
|
|
|
|
/***/ },
|
|
/* 6 */
|
|
/***/ function(module, exports, __webpack_require__) {
|
|
|
|
/* WEBPACK VAR INJECTION */(function(module) {// import
|
|
var Layer = __webpack_require__(3)
|
|
, Network = __webpack_require__(4)
|
|
, Trainer = __webpack_require__(5)
|
|
|
|
/*******************************************************************************************
|
|
ARCHITECT
|
|
*******************************************************************************************/
|
|
|
|
// Collection of useful built-in architectures
|
|
var Architect = {
|
|
|
|
// Multilayer Perceptron
|
|
Perceptron: function Perceptron() {
|
|
|
|
var args = Array.prototype.slice.call(arguments); // convert arguments to Array
|
|
if (args.length < 3)
|
|
throw new Error("not enough layers (minimum 3) !!");
|
|
|
|
var inputs = args.shift(); // first argument
|
|
var outputs = args.pop(); // last argument
|
|
var layers = args; // all the arguments in the middle
|
|
|
|
var input = new Layer(inputs);
|
|
var hidden = [];
|
|
var output = new Layer(outputs);
|
|
|
|
var previous = input;
|
|
|
|
// generate hidden layers
|
|
for (var level in layers) {
|
|
var size = layers[level];
|
|
var layer = new Layer(size);
|
|
hidden.push(layer);
|
|
previous.project(layer);
|
|
previous = layer;
|
|
}
|
|
previous.project(output);
|
|
|
|
// set layers of the neural network
|
|
this.set({
|
|
input: input,
|
|
hidden: hidden,
|
|
output: output
|
|
});
|
|
|
|
// trainer for the network
|
|
this.trainer = new Trainer(this);
|
|
},
|
|
|
|
// Multilayer Long Short-Term Memory
|
|
LSTM: function LSTM() {
|
|
|
|
var args = Array.prototype.slice.call(arguments); // convert arguments to array
|
|
if (args.length < 3)
|
|
throw new Error("not enough layers (minimum 3) !!");
|
|
|
|
var last = args.pop();
|
|
var option = {
|
|
peepholes: Layer.connectionType.ALL_TO_ALL,
|
|
hiddenToHidden: false,
|
|
outputToHidden: false,
|
|
outputToGates: false,
|
|
inputToOutput: true,
|
|
};
|
|
if (typeof last != 'number') {
|
|
var outputs = args.pop();
|
|
if (last.hasOwnProperty('peepholes'))
|
|
option.peepholes = last.peepholes;
|
|
if (last.hasOwnProperty('hiddenToHidden'))
|
|
option.hiddenToHidden = last.hiddenToHidden;
|
|
if (last.hasOwnProperty('outputToHidden'))
|
|
option.outputToHidden = last.outputToHidden;
|
|
if (last.hasOwnProperty('outputToGates'))
|
|
option.outputToGates = last.outputToGates;
|
|
if (last.hasOwnProperty('inputToOutput'))
|
|
option.inputToOutput = last.inputToOutput;
|
|
} else
|
|
var outputs = last;
|
|
|
|
var inputs = args.shift();
|
|
var layers = args;
|
|
|
|
var inputLayer = new Layer(inputs);
|
|
var hiddenLayers = [];
|
|
var outputLayer = new Layer(outputs);
|
|
|
|
var previous = null;
|
|
|
|
// generate layers
|
|
for (var layer in layers) {
|
|
// generate memory blocks (memory cell and respective gates)
|
|
var size = layers[layer];
|
|
|
|
var inputGate = new Layer(size).set({
|
|
bias: 1
|
|
});
|
|
var forgetGate = new Layer(size).set({
|
|
bias: 1
|
|
});
|
|
var memoryCell = new Layer(size);
|
|
var outputGate = new Layer(size).set({
|
|
bias: 1
|
|
});
|
|
|
|
hiddenLayers.push(inputGate);
|
|
hiddenLayers.push(forgetGate);
|
|
hiddenLayers.push(memoryCell);
|
|
hiddenLayers.push(outputGate);
|
|
|
|
// connections from input layer
|
|
var input = inputLayer.project(memoryCell);
|
|
inputLayer.project(inputGate);
|
|
inputLayer.project(forgetGate);
|
|
inputLayer.project(outputGate);
|
|
|
|
// connections from previous memory-block layer to this one
|
|
if (previous != null) {
|
|
var cell = previous.project(memoryCell);
|
|
previous.project(inputGate);
|
|
previous.project(forgetGate);
|
|
previous.project(outputGate);
|
|
}
|
|
|
|
// connections from memory cell
|
|
var output = memoryCell.project(outputLayer);
|
|
|
|
// self-connection
|
|
var self = memoryCell.project(memoryCell);
|
|
|
|
// hidden to hidden recurrent connection
|
|
if (option.hiddenToHidden)
|
|
memoryCell.project(memoryCell, Layer.connectionType.ALL_TO_ELSE);
|
|
|
|
// out to hidden recurrent connection
|
|
if (option.outputToHidden)
|
|
outputLayer.project(memoryCell);
|
|
|
|
// out to gates recurrent connection
|
|
if (option.outputToGates) {
|
|
outputLayer.project(inputGate);
|
|
outputLayer.project(outputGate);
|
|
outputLayer.project(forgetGate);
|
|
}
|
|
|
|
// peepholes
|
|
memoryCell.project(inputGate, option.peepholes);
|
|
memoryCell.project(forgetGate, option.peepholes);
|
|
memoryCell.project(outputGate, option.peepholes);
|
|
|
|
// gates
|
|
inputGate.gate(input, Layer.gateType.INPUT);
|
|
forgetGate.gate(self, Layer.gateType.ONE_TO_ONE);
|
|
outputGate.gate(output, Layer.gateType.OUTPUT);
|
|
if (previous != null)
|
|
inputGate.gate(cell, Layer.gateType.INPUT);
|
|
|
|
previous = memoryCell;
|
|
}
|
|
|
|
// input to output direct connection
|
|
if (option.inputToOutput)
|
|
inputLayer.project(outputLayer);
|
|
|
|
// set the layers of the neural network
|
|
this.set({
|
|
input: inputLayer,
|
|
hidden: hiddenLayers,
|
|
output: outputLayer
|
|
});
|
|
|
|
// trainer
|
|
this.trainer = new Trainer(this);
|
|
},
|
|
|
|
// Liquid State Machine
|
|
Liquid: function Liquid(inputs, hidden, outputs, connections, gates) {
|
|
|
|
// create layers
|
|
var inputLayer = new Layer(inputs);
|
|
var hiddenLayer = new Layer(hidden);
|
|
var outputLayer = new Layer(outputs);
|
|
|
|
// make connections and gates randomly among the neurons
|
|
var neurons = hiddenLayer.neurons();
|
|
var connectionList = [];
|
|
|
|
for (var i = 0; i < connections; i++) {
|
|
// connect two random neurons
|
|
var from = Math.random() * neurons.length | 0;
|
|
var to = Math.random() * neurons.length | 0;
|
|
var connection = neurons[from].project(neurons[to]);
|
|
connectionList.push(connection);
|
|
}
|
|
|
|
for (var j = 0; j < gates; j++) {
|
|
// pick a random gater neuron
|
|
var gater = Math.random() * neurons.length | 0;
|
|
// pick a random connection to gate
|
|
var connection = Math.random() * connectionList.length | 0;
|
|
// let the gater gate the connection
|
|
neurons[gater].gate(connectionList[connection]);
|
|
}
|
|
|
|
// connect the layers
|
|
inputLayer.project(hiddenLayer);
|
|
hiddenLayer.project(outputLayer);
|
|
|
|
// set the layers of the network
|
|
this.set({
|
|
input: inputLayer,
|
|
hidden: [hiddenLayer],
|
|
output: outputLayer
|
|
});
|
|
|
|
// trainer
|
|
this.trainer = new Trainer(this);
|
|
},
|
|
|
|
Hopfield: function Hopfield(size) {
|
|
|
|
var inputLayer = new Layer(size);
|
|
var outputLayer = new Layer(size);
|
|
|
|
inputLayer.project(outputLayer, Layer.connectionType.ALL_TO_ALL);
|
|
|
|
this.set({
|
|
input: inputLayer,
|
|
hidden: [],
|
|
output: outputLayer
|
|
});
|
|
|
|
var trainer = new Trainer(this);
|
|
|
|
var proto = Architect.Hopfield.prototype;
|
|
|
|
proto.learn = proto.learn || function(patterns)
|
|
{
|
|
var set = [];
|
|
for (var p in patterns)
|
|
set.push({
|
|
input: patterns[p],
|
|
output: patterns[p]
|
|
});
|
|
|
|
return trainer.train(set, {
|
|
iterations: 500000,
|
|
error: .00005,
|
|
rate: 1
|
|
});
|
|
};
|
|
|
|
proto.feed = proto.feed || function(pattern)
|
|
{
|
|
var output = this.activate(pattern);
|
|
|
|
var pattern = [];
|
|
for (var i in output)
|
|
pattern[i] = output[i] > .5 ? 1 : 0;
|
|
|
|
return pattern;
|
|
}
|
|
}
|
|
}
|
|
|
|
// Extend prototype chain (so every architectures is an instance of Network)
|
|
for (var architecture in Architect) {
|
|
Architect[architecture].prototype = new Network();
|
|
Architect[architecture].prototype.constructor = Architect[architecture];
|
|
}
|
|
|
|
// export
|
|
if (module) module.exports = Architect;
|
|
/* WEBPACK VAR INJECTION */}.call(exports, __webpack_require__(2)(module)))
|
|
|
|
/***/ }
|
|
/******/ ]); |