Rate option - what is it eaten with?

Question

Neural networks

There are many different ready-made neural networks for Node.js on Github. For example. In the network learning settings options, there is such a parameter as rate - the learning rate.
I just can’t understand the speed of 0.1 greater than 0.3? And how do 0.3 and 0.03 relate? What speed is higher?
Experiments somehow did not reveal a clear result. I think it's the same.)

Answer 1

Read about how gradient descent works .
In the learning process, you minimize the loss function, look for its bottom:
“Learning rate” == “step size”. If the steps are too big, there is a risk of missing your stop. If it is too small, you will come, but walking is faster. :)
The illustration shows an even more optimal strategy, when the step size changes with each iteration from large to small.