Created
June 20, 2018 17:45
-
-
Save curiousily/2a8fb08d0e20961256f5a88cb7b8e10d to your computer and use it in GitHub Desktop.
Create a Simple Linear Regression model in TensorFlow.js that given some number from the Fibonacci sequence predicts the next one while only running in the browser!
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
// What is a Tensor? | |
const myFirstTensor = tf.scalar(42) | |
console.log(myFirstTensor) | |
myFirstTensor.print() | |
const oneDimTensor = tf.tensor1d([1, 2, 3]) | |
oneDimTensor.print() | |
// Preparing the training data | |
function fibonacci(num){ | |
var a = 1, b = 0, temp; | |
var seq = [] | |
while (num > 0){ | |
temp = a; | |
a = a + b; | |
b = temp; | |
seq.push(b) | |
num--; | |
} | |
return seq; | |
} | |
const fibs = fibonacci(100) | |
const xs = tf.tensor1d(fibs.slice(0, fibs.length - 1)) | |
const ys = tf.tensor1d(fibs.slice(1)) | |
const xmin = xs.min(); | |
const xmax = xs.max(); | |
const xrange = xmax.sub(xmin); | |
function norm(x) { | |
return x.sub(xmin).div(xrange); | |
} | |
xsNorm = norm(xs) | |
ysNorm = norm(ys) | |
// Building our model | |
const a = tf.variable(tf.scalar(Math.random())) | |
const b = tf.variable(tf.scalar(Math.random())) | |
a.print() | |
b.print() | |
function predict(x) { | |
return tf.tidy(() => { | |
return a.mul(x).add(b) | |
}); | |
} | |
// Training | |
function loss(predictions, labels) { | |
return predictions.sub(labels).square().mean(); | |
} | |
const learningRate = 0.5; | |
const optimizer = tf.train.sgd(learningRate); | |
const numIterations = 10000; | |
const errors = [] | |
for (let iter = 0; iter < numIterations; iter++) { | |
optimizer.minimize(() => { | |
const predsYs = predict(xsNorm); | |
const e = loss(predsYs, ysNorm); | |
errors.push(e.dataSync()) | |
return e | |
}); | |
} | |
// Making predictions | |
console.log(errors[0]) | |
console.log(errors[numIterations - 1]) | |
xTest = tf.tensor1d([2, 354224848179262000000]) | |
predict(xTest).print() | |
a.print() | |
b.print() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment