I recently interviewed (telephonic) at a large database software company for an ML in software engineering position in India. Here are the questions asked:

1. Why use CNN? What advantages they have over using a normal NN?

2. What is Linear Independence of vectors?

3. How does one validate the predictions are correct?

4. What is bias and variance?

5. What is overfitting? What are the methods to prevent overfitting? How does one decide what features to use? What is correlation?

6. What is gradient descent? Why use it?

7. What is the difference between linear regression and logistic regression? Where are they applied? Why are they called regression methods?

8. What is regularization? What are the various types?

9. When is a sigmoid function used?

10. Is tanh function a non linear function? What is difference between tanh & sigmoid? When is it used?

11. What other non linear function have you used? What is ReLU & Leaky ReLU? Why use Leaky ReLU? When are each of them used?

12. What are the various types of loss functions?