Using the Backpropagation algorithm, train the two$-$layer neural network with the following structure in such a way that it can handle data

Learn a tutorial similar to the one shown in the figure below.

$$

The program is written using the Python programming language.

$y=tanh(s)$

$y^{\text{'}}=1-tan{h}^2(s)$

The function of nerve activity is sigmoid.

To generate data, use the random number generation function with Gaussian distribution $($normal$).$ The desired output of the network is

Consider $1$ for blue data and $1$ for red data.

The number $1$ of the number of hidden layer nerves should be received as input from the user.

Implement all three versions of gradient descent algorithm $($BGD$,$ MBGD and SGD$).$

In each training step, show the dividing line of the nerves of the first layer compared to the training data

that the convergence of the network output to the training data is completely clear.

In each step of training, calculate the error of the network on the entire data based on the MSE criterion and show it in a graph

give