Adjust the training parameter and the number of training epoch so the NN converges satisfactorily. Use a training of 80% of the data and a validation set of 20% of the data. Represent the experiments and interpret the results. import numpy as np def data(name): \#xor and circle problems are generated here. sigma =0.6 \# standard deviation of the clusters classes =[0,1,1,0] X=sigma inp.random. randn (2,4N ) \# Generate 4N random vectors (gaussian) centered around zero mean=np.array ([[1,1,1,1],[1,1,1,1]]) \# define four means M=np.ones ((N,2)) mean [:,0]# Means of the first cluster y=np. ones ((1,N))cla for i in range (1,4): m=np.ones ((N,2)) mean [:,i] \# Means of cluster i M=np. concatenate ((M,m)) \#Concatenate all means y=np. concatenate ((y,np.ones ((1,N)) classes[i] , axis=1) \# concatenate labels X=X+M \# Add means to the data. The data can be represented as follows (left: xor problem; right: circle problem)