Last active
February 25, 2019 20:27
-
-
Save LlamaTrauma/c70a197c123d4328cae9c71b6f08e41d to your computer and use it in GitHub Desktop.
Broken JavaScript Neural Network
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
const math = require('mathjs'); | |
var fs = require('fs'); | |
//These files are found here: http://yann.lecun.com/exdb/mnist/ | |
var imageFile = fs.readFileSync(".../train-images-idx3-ubyte/train-images.idx3-ubyte"); | |
var labelFile = fs.readFileSync(".../train-labels-idx1-ubyte/train-labels.idx1-ubyte"); | |
var pixelValues = []; | |
for (var image = 0; image <= 6000; image++) { | |
var pixels = []; | |
for (var y = 0; y <= 27; y++) { | |
for (var x = 0; x <= 27; x++) { | |
pixels.push(imageFile[(image * 28 * 28) + (x + (y * 28)) + 16]); | |
} | |
} | |
pixelValues.push(pixels); | |
} | |
var labelValues = []; | |
for (var image = 0; image <= 6000; image++) { | |
var out = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0]; | |
out[labelFile[image + 8]] = 1; | |
labelValues.push(out); | |
} | |
function getMostLikely(a){ | |
var best = 0; | |
var index = 0; | |
for(var i = 0; i < a.length; i ++){ | |
if(best < a[i]){ | |
best = a[i]; | |
index = i; | |
} | |
} | |
return(index); | |
}; | |
const e = 2.71828182845904523; | |
function sigmoid(a){ | |
return(1 / (1 + math.pow(e, a * -1))); | |
} | |
function sigmoidDerivitive(a){ | |
return(a * (1 - a)); | |
} | |
function squareDiff(a, b){ | |
var out = []; | |
for(var i = 0; i < a._data.length; i ++){ | |
for(var j = 0; j < a._data[i].length; j ++){ | |
out.push(math.pow(a._data[i][j] - b._data[i][j], 2)); | |
} | |
} | |
var outy = 0; | |
for(var i = 0; i < out.length; i ++){ | |
outy += out[i]; | |
} | |
outy /= out.length; | |
return(outy); | |
} | |
function multMatrices(a, b){ | |
var out = new math.matrix(); | |
out.resize(a._size, 0); | |
for(var i = 0; i < a._data.length; i ++){ | |
for(var j = 0; j < a._data[i].length; j ++){ | |
out._data[i][j] = a._data[i][j] * b._data[i][j]; | |
} | |
} | |
return(out); | |
}; | |
function transposeMatrix(a){ | |
var out = new math.matrix(); | |
out.resize([a._size[1], a._size[0]]); | |
for(var i = 0; i < a._data.length; i ++){ | |
for(var j = 0; j < a._data[i].length; j ++){ | |
out._data[j][i] = a._data[i][j]; | |
} | |
} | |
return(out); | |
}; | |
function setAll(a, b){ | |
var out = new math.matrix(); | |
out.resize(a._size); | |
function setit(c){ | |
var outy; | |
if(b === "random"){ | |
outy = math.random(0, 1); | |
} | |
if(b === "randomlow"){ | |
outy = math.random(0, 0.05); | |
} | |
else if(b === "sigmoid"){ | |
outy = sigmoid(c); | |
} | |
else if(b === "sigmoidDerivitive"){ | |
outy = (sigmoidDerivitive(c)); | |
} | |
else{ | |
outy = c * b; | |
} | |
return(outy); | |
}; | |
for(var i = 0; i < a._data.length; i ++){ | |
if(a._size.length < 2){ | |
out._data[i] = setit(a._data[i]); | |
} | |
else{ | |
for(var j = 0; j < a._data[i].length; j ++){ | |
out._data[i][j] = setit(a._data[i][j]); | |
} | |
} | |
} | |
return(out); | |
}; | |
function NeuralNetwork(x, y){ | |
this.input = x; | |
this.y = y; | |
this.sizes = [this.input._size[1], 300, this.y._size[1]]; | |
this.layers = this.sizes.length - 1; | |
this.lyrs = [this.input]; | |
this.weights = []; | |
this.dweights = []; | |
for(var i = 0; i < this.layers; i ++){ | |
this.weights.push(new math.matrix()); | |
this.weights[i].resize([this.sizes[i], this.sizes[i + 1]]); | |
this.weights[i] = setAll(this.weights[i], "randomlow"); | |
} | |
this.output = new math.matrix(); | |
this.output.resize(this.y._size); | |
}; | |
NeuralNetwork.prototype.set = function(x, y){ | |
this.input = x; | |
this.lyrs = [this.input]; | |
this.y = y; | |
}; | |
NeuralNetwork.prototype.feedforward = function(){ | |
for(var i = 0; i < this.weights.length; i ++){ | |
this.lyrs[i + 1] = math.multiply(this.lyrs[i], this.weights[i]); | |
this.lyrs[i + 1] = setAll(this.lyrs[i + 1], "sigmoid"); | |
} | |
this.output = this.lyrs[this.lyrs.length - 1]; | |
}; | |
NeuralNetwork.prototype.backpropogate = function(){ | |
this.antis = [ | |
function(a, b, c){ | |
return( | |
math.multiply(transposeMatrix(a[0]), multMatrices(math.multiply(multMatrices(math.multiply(math.subtract(b.y, b.output), 2), setAll(b.output, "sigmoidDerivitive")), transposeMatrix(c)), setAll(a[1], "sigmoidDerivitive"))) | |
); | |
}, | |
function(a, b, c){ | |
return( | |
math.multiply(transposeMatrix(a[0]), multMatrices(math.multiply(math.subtract(b.y, b.output), 2), setAll(b.output, "sigmoidDerivitive"))) | |
); | |
}]; | |
this.input = []; | |
this.weightInput = 0; | |
for(var i = this.weights.length - 1; i >= 0; --i){ | |
this.input.unshift(this.lyrs[i]); | |
this.weightInput = (i === this.weights.length - 1 ? 0 : this.weights[i + 1]); | |
this.dweights[i] = this.antis[i](this.input, this, this.weightInput); | |
} | |
for(var i = 0; i < this.dweights.length; i ++){ | |
this.weights[i] = math.add(this.weights[i], this.dweights[i]); | |
} | |
}; | |
var feed = new math.matrix([pixelValues[0]]); | |
var feed2 = new math.matrix([labelValues[0]]); | |
var t = new NeuralNetwork(feed, feed2); | |
t.feedforward(); | |
var to = 270; | |
for(var i = 0; i < to + 1; i ++){ | |
feed = new math.matrix([pixelValues[i]]); | |
feed2 = new math.matrix([labelValues[i]]); | |
t.set(feed, feed2); | |
t.feedforward(); | |
t.backpropogate(); | |
console.log("Loss: " + squareDiff(t.y, t.output) + " " + i); | |
if(i === to){ | |
console.log(getMostLikely(t.output._data[0])); | |
console.log(t.output._data); | |
} | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment