Merge pull request #1 from will-newmarch/release-0.2.0

Release 0.2.0

.gitignore

@@ -61,3 +61,5 @@ typings/
 
 # next.js build output
 .next
+
+.idea

.travis.yml

@@ -0,0 +1,3 @@
+language: node_js
+node_js:
+  - "8"

README.md

@@ -1,5 +1,6 @@
+[![Build Status](https://travis-ci.org/will-newmarch/intuitive-neural-network.svg?branch=master)](https://travis-ci.org/will-newmarch/intuitive-neural-network)
 # Intuitive Neural Network
-### When would like to understand what's going on without the mathematics!
+### When you would like to understand what's going on without the mathematics!
 A JavaScript based, object-orientated approach to a Neural Network library.
 
 The two aims behind creating this library are:
@@ -9,3 +10,124 @@ The two aims behind creating this library are:
 ---
 
 Run the XOR example with `npm test`
+
+---
+
+## A Simple Implementation (XOR Problem)
+
+    // Build the network...
+    var network = new Network({
+        layers: [2,2,1],
+        bias: false
+    });
+
+    // Training data (x in, y out)
+    var data = [
+        {x: [0,0], y: [0]},
+        {x: [0,1], y: [1]},
+        {x: [1,0], y: [1]},
+        {x: [1,1], y: [0]}
+    ];
+
+    // Training the network...
+    var epochs = 10000;
+    var learningRate = 0.01;
+
+    for (var h = 0; h < epochs; h++) {
+
+        for (var i = 0; i < data.length; i++) {
+
+            let index = Math.floor(Math.random() * data.length);
+            
+            network
+                .fire(data[index].x)
+                .backPropagate(data[index].y)
+                .applyError(learningRate)
+                .reset();
+
+        }
+    }
+    // Done.
+
+    // Testing the trained network...
+    for(var i = 0; i < data.length; i++) {
+
+        network.fire(data[i].x);
+
+        var activation = network.layers[network.layers.length-1].neurons[0].activation;
+
+        // expect Math.round(activation) to equal data[i].y[0]
+
+        network.reset();
+
+    }
+    // Done.
+
+---
+
+
+<a name="Network"></a>
+
+## Network
+**Kind**: global class  
+
+* [Network](#Network)
+    * [new Network(settings)](#new_Network_new)
+    * [.fire(signals)](#Network+fire) ⇒
+    * [.backPropagate(errors)](#Network+backPropagate) ⇒
+    * [.applyError(learningRate)](#Network+applyError) ⇒
+    * [.reset()](#Network+reset) ⇒
+
+<a name="new_Network_new"></a>
+
+### new Network(settings)
+Constructor for Network
+
+
+| Param | Type |
+| --- | --- |
+| settings | <code>object</code> | 
+
+<a name="Network+fire"></a>
+
+### network.fire(signals) ⇒
+Fire the input layer's Neurons with supplied array of floats
+
+**Kind**: instance method of [<code>Network</code>](#Network)  
+**Returns**: Network (for chaining purposes)  
+
+| Param | Type |
+| --- | --- |
+| signals | <code>array</code> | 
+
+<a name="Network+backPropagate"></a>
+
+### network.backPropagate(errors) ⇒
+Initialise back propagation through network with supplied array of floats
+
+**Kind**: instance method of [<code>Network</code>](#Network)  
+**Returns**: Network (for chaining purposes)  
+
+| Param | Type |
+| --- | --- |
+| errors | <code>array</code> | 
+
+<a name="Network+applyError"></a>
+
+### network.applyError(learningRate) ⇒
+Trigger each synapse to apply its error to its weight
+
+**Kind**: instance method of [<code>Network</code>](#Network)  
+**Returns**: Network (for chaining purposes)  
+
+| Param | Type |
+| --- | --- |
+| learningRate | <code>float</code> | 
+
+<a name="Network+reset"></a>
+
+### network.reset() ⇒
+Reset all the Neurons and Synapses back to their initial state
+
+**Kind**: instance method of [<code>Network</code>](#Network)  
+**Returns**: Network (for chaining purposes)  

src/Network.js

@@ -98,6 +98,7 @@ class Network {
 	/**
 	 * Fire the input layer's Neurons with supplied array of floats
 	 * @param {array} signals 
+	 * @returns Network (for chaining purposes)
 	 */
 	fire(signals) {
 		for (var i = 0; i < this.layers[0].neurons.length; i++) {
@@ -114,6 +115,7 @@ class Network {
 	/**
 	 * Initialise back propagation through network with supplied array of floats
 	 * @param {array} errors 
+	 * @returns Network (for chaining purposes)
 	 */
 	backPropagate(errors) {
 		for (var i = 0; i < errors.length; i++) {
@@ -125,6 +127,7 @@ class Network {
 	/**
 	 * Trigger each synapse to apply its error to its weight
 	 * @param {float} learningRate 
+	 * @returns Network (for chaining purposes)
 	 */
 	applyError(learningRate) {
 		this.layers.map(l => {
@@ -136,6 +139,7 @@ class Network {
 				}
 			});
 		});
+		return this;
 	}
 
 	/**
@@ -153,6 +157,7 @@ class Network {
 
 	/**
 	 * Reset all the Neurons and Synapses back to their initial state
+	 * @returns Network (for chaining purposes)
 	 */
 	reset() {
 		this.layers.map(l => {
@@ -165,6 +170,7 @@ class Network {
 				}
 			});
 		});
+		return this;
 	}
 
 	/**

xor-problem.test.js

@@ -2,16 +2,9 @@ const Network = require('./src/Network.js');
 
 test('library solves XOR problem', () => {
 
-    // Various settings...
-    var learningRate = 0.01;
-    var epochs = 100000;
-    var activation = 'sigmoid';
-
     // Build the network...
     var network = new Network({
         layers: [2,2,1],
-        hiddenActivationType: activation,
-        outputActivationType: 'identity',
         bias: false
     });
 
@@ -22,29 +15,27 @@ test('library solves XOR problem', () => {
         {x: [1,1], y: [0]}
     ];
 
-    // Training...
+    // Training the network...
+    var epochs = 10000;
+    var learningRate = 0.01;
 
     for (var h = 0; h < epochs; h++) {
 
         for (var i = 0; i < data.length; i++) {
 
             let index = Math.floor(Math.random() * data.length);
 
-            network.fire(data[index].x);
-
-            network.backPropagate(data[index].y);
-
-            network.applyError(learningRate);
-
-            network.reset();
+            network
+                .fire(data[index].x)
+                .backPropagate(data[index].y)
+                .applyError(learningRate)
+                .reset();
 
         }
     }
-
     // Done.
 
-    // Testing...
-
+    // Testing the trained network...
     for(var i = 0; i < data.length; i++) {
 
         network.fire(data[i].x);
@@ -56,7 +47,6 @@ test('library solves XOR problem', () => {
         network.reset();
 
     }
-
     // Done.
 
 });