$24
Submit your answers here: https://forms.gle/QuY3GBfJoAkbT9FA7
Q1. Complete the following code for a NN based classifier for iris data. Use only numpy
library, and implement other functions yourself. (35 marks) https://drive.google.com/open?id=1EBrahr2ZVXcbETRsyuri2b78cS1tCA4n
Note: It will be automatically evaluated by importing the functions you have written. So please use
if __name__=="__main__":
for lines which are not a part of the function.
Q2.
(20 marks)
a. Solve the Q1 using keras library. Use
i. sigmoid and mean squared error loss
ii. softmax and categorical cross-entropy loss
Print confusion matrices and plot loss function vs number of epochs in your ipynb.
b. Which of the above two schemes gives better results and why? (5 marks)