Instructor

Prof. Amit Sethi

Semester

Autumn ‘20

Course Difficulty

Intermediate

Time Commitment Required

Apart from the lecture hours, giving time to the tutorials regularly was usually sufficient. Moderate commitment is required. For the weeks that have the programing assignment submission, more efforts are expected as assignments take you through a steep learning curve specially if this is your first time coding DL.

Grading Policy and Statistics

Prof. Amit stated in the beginning itself that grading is going to be absolute. 100+ would yield AP, 90+ would yield AA, 80+ would yield AB and so on…

Given the extra bonus marks in the Project, Mid-Semester, End-Semester and the assignments, one can easily score 90+ given he/she is regular. The course would be moderately difficult in general and for the ones who have been exposed to the deep learning world, the course would be relatively on the easier side.

Attendance Policy

There was no attendance policy followed.

Pre-requisites

An Intro to ML course from any department would suffice. At our time, Prof. Amit also conducted a quiz on the content of Intro to ML to let students take a call on whether they are ready to take the Adv. ML course or not.

Evaluation Scheme

Weekly Moodle Quizzes
Midsem
Endsem
Assignments (2)
Project

Topics Covered in the Course

Supervised Machine Learning, Dense Neural Networks, Backpropagation, Convolutional Neural Networks, Recurrent Neural Networks, LSTM/GRU, Loss functions and regularization techniques, Gradient Descent for non-convex optimization, Training deep networks, GANs, Self-Supervised Learning, Semi-Supervised Learning, Basic NLP models, Attention Mechanism in NLP

Teaching Style

Videos of Prof. Amit Sethi(EE782, Autumn ‘19) were available on CDEEP. He usually expected the students to have watched the lectures on CDEEP prior to the class and would then give a brief review of the content and address any doubts. Towards the later half of the course, he switched to taking live lectures.

Tutorials/Assignments/Projects

The tutorials were regularly discussed by the Prof. and played an important role in understanding the content being taught in the course. The Assignments and Projects were more on the programming and application side and would really push one to grasp good knowledge of libraries like PyTorch. A good thing was also that the professor’s intent was to make the students learn through the journey of solving assignments and doing projects, so he never evaluated the assignments/projects based on the accuracies your model achieved. The hard work put in and the knowledge gain you demonstrate was always counted for even if things don’t seem to work well in code.

Feedback on Exams

Quizzes: There were weekly quizzes conducted on moodle. They were usually MCQs with possibly multiple answers correct and one was also required to type out the reasons validating his/her choice. The quizzes usually tested the material discussed in the weekly tutorials and would range from easy to moderately difficult in terms of scoring.

Midsem/Endsem: The midsem and endsem were open notes and had open ended questions to a large extent. For instance, for the endsem exam, Prof. had shared a list of several research papers and the questions were mainly based on the understanding and applications of those papers. The exams of this course looked very relevant from the point of view of understanding concepts in detail because one would have no choice apart from understanding and digesting well the contents of the papers.

Course Importance

The knowledge and the experience gained throughout the journey of this course could help you immensely in performing well in internships relevant to this field as apart from just the theoretical concepts, the course helps you develop other skills like good coding and a habit of reading research papers.

Review By: Aishwarya Agarwal