Information$$ Data provided: dataset.csv

$$ Classes are in the last column

$$ Use K$-$fold $=5$

$$ Design using tensorflow and keras

Design an MLP$-$Perceptron and classify the assigned dataset using K$-$fold$=5$:

a$)$ Learning rate: $0\mathrm{.}1$

b$)$ Number of hidden layers: $4$a$.$ Number of neurons per layer: $50$b$.$ Activation function: Relu

c$)$ Hyperparameterse. optimizer $=$ SGDb$.$ loss $=$ 'mse'c. metrics $=$ accuracyd. batch$\_$size $=64$e$.$ Epoch $=5$

d$)$ Plot Accuracy vs Epoch $($X$-$axis: Epoch, Y$-$axis: Accuracy$)$e$)$ Perform the test, that is$,$ the classification with the test data and print the accuracy

Now change the layout as follows

a$)$ Learning rate: $0\mathrm{.}01$

b$)$ Number of hidden layers: $6$a$.$ Number of neurons per layer: $60$b$.$ Activation function: Relu

c$)$ Hyperparameterse. optimizer $=$ Adamb. loss $=$ 'binary$\_$crossentropy'c. metrics $=$ 'accuracy'd. batch$\_$size $=128$e$.$ Epoch $=5$

d$)$ Plot Accuracy versus Epoch $($X$-$axis: Epoch, Y$-$axis: Accuracy$)$e$)$ Perform the test after training and print the accuracy

Answer the following questions:

$1.$ Which is the best model, i$.$e$.$ exercise$-1$ or $2?$ explain and present your arguments for your choice

$2.$ Which of the training hyperparameters do you think has the most influence on accuracy? Why? List your arguments to defend your point.

$3.$ Write down all the changes that have surprised you in this lab