Why don't TensorFlow sometimes normalize data by one?

Question

JavaScript

Occasionally I stumble upon similar examples of using TensorFlow: https://medium.com/tensorflow/getting-started-with...
where data is not normalized before being thrown into TensorFlow:

const xs = tf.tensor2d([-1, 0, 1, 2, 3, 4], [6, 1]);
const ys = tf.tensor2d([-3, -1, 1, 3, 5, 7], [6, 1]);
await model.fit(xs, ys, {epochs: 500});

And it still works fine :) Shouldn't the data be pre-normalized by one?

Answer 1

Data normalization is just one way to speed up learning. So that it descends faster instead of fluctuating left and right, relatively speaking. This is an optimization, not part of the algorithm.

Answer 2

It seems not. The backpropagation method avoids this.