The input data $1,x2,$ y can be loaded from file: $"$x$1\_$x$2\_$y$\_$circle$2.$csv$"$ Work off this file to implement the back$-$propagation gradient for this classifier. Split the

data so that you train on $80\%$ and test with the remaining $20\%.$ Submit the following:

Code $($with algorithm derived in class$)$

Plot of converging cost function $J$ and final value $($train data$)$

Plots of classification showing $\text{'}x\text{'}$s$,$ o$\text{'}$s$,$ and decision boundary. Do this for both training and testing data.

Compare the accuracies of training vs testing. Accuracy can be calcuated with this command: acc$=np*mean(np*r{\displaystyle \underset{\phantom{}}{\overset{\phantom{}}{}}}(A)==Y{)}^{***}100$

Extra credit: $15\%$ on separate notebook file using Keras