$2\mathrm{.}1$ We have seen that there are quite a few factors that necessitate development of technqiues for

Explanability of Black Box Machine Learning models, Which of the requirements stated below does NOT

constitute such a factor

a$)$ end$-$user requirements to relate specific observed betavior of outputs to specific inputs

b$)$ legal requirements for commercialization of products

c$)$ compulsion to attain higher accuracies of solution

d$)$ gaining trust and confidence of customers towards commercialized products.

$2\mathrm{.}2.$ Let's say that you need to train an algorithm for classifying objects and you have access to a small

amount of data. You also have a pre$-$trained model which was trained for a similar $($but not same$)$ task.

What would you do$?$

a$)$ Train a deep Convolutional Neural Network

b$)$ Evaluate the important layers of the pre$-$trained model and re$-$tune them

c$)$ Re$-$train only the last layer of the pre$-$trained model

d$)$ Re$-$train only the first layer of the pre$-$trained model.