Is it worth shuffling data to train a neural network during training?

Question

Python

Somewhere I briefly read that for faster convergence, it is desirable to shuffle the input data for the neural network during the learning process. Where can you read about it? Is the game worth the candle?
The second question is, if it is beneficial for the tasks set, then how to do it right?

So?

It was like this:
[0, 0, 1, 1]
[0, 1, 0, 1]
[1, 1, 0, 0]
Shuffle, it became like this:
[0, 1, 0, 1]
[1, 1, 0 , 0]
[0, 0, 1, 1
] shuffled the cases

Or so?

Было так:
[0, 0, 1, 1]
[0, 1, 0, 1]
[1, 1, 0, 0]
Тасуем, стало так:
[0, 1, 1, 0]
[1, 0, 1, 0]
[1, 0, 0, 1]
Т.е. перетасовали значения в каждом случае

Answer 1

The point is that for each stage you make a random sample from the original dataset. This speeds up the process and avoids overfitting. Read "Deep learning. Dive into the world of neural networks".