Project 2

Q.Write steps for gradient descent for 2 input features and output having 3 classes.

1.Loss function = -[ yloga1 + y2loga2 + y3loga3 + y4loga4]

2.We have to found derivatives for 0th layer with respect to z

3.We have to found derivatives for 1st hidden layer by applying any activation function

4.As we know we have 3 classes in output

5.So applying softmax activation function to output layer

6.We will get the highest value out of 3 classes and the datapoint will be considered to be of that class