Exchange Diaries  

Siddharth Gupta | Electronics and Electrical Communication Engg.

Yale University, USA

Research Area: Domain Adaptation in Machine Learning using Autoencoder structure

From where did you get the motivation? How did it all start? How did you handle your preparation with your academics?

There was no plan, however, because of this COVID pandemic. We went home and had a lot of time and that International Relations Cell was providing these work from home internships. So I applied to the few of those. I wanted to do something in machine learning, as I had already done some online courses. Plus I wanted to do a hands-on project.
No, there was no stress of academics that was added on to this. As I already mentioned, I did my remote internship during this COVID pandemic, so the college wasn’t working as such, and all I had to focus on was on this internship.

What was the role of CG?

If you are applying for programs through IIT Kharagpur, then I think CG plays a vital role. As they don’t have any means of testing one’s knowledge in the field, thus CG, SOP, and the introduction that you have written about yourself remain as the criteria for selection. In all, if you are going through the institute, then I feel that CG matters a lot.

How is your university going to manage everything in this confusion of Covid-19?

The international relations cell managed quite well. I had applied through the ft portal, so it was very well coordinated and organised.
In the university point of view, I contacted the professor. We were around 3-4 people at the start of the project. We initially had some meetings over video calls on zoom. There he explained the topic and asked if we had to continue it or not. Every week we had video calls where we had to present all of the works that we had done. Basically what I think is that it was mostly a software part of the thing so the things going offline or online didn’t matter much.

What were the difficulties you faced during the entire process? Did the current pandemic situation affect you in any way?

Due to this online stuff, I had to take prior permission to have a separate video call with the professors regarding the doubts. So in that sense, there was some difficulty in communication. But overall it was a pleasant experience. The professor was even excellent. He interacted with us a lot.

What are your plans and ambitions?

I’m still not sure about my future endeavours. I just had the opportunity in summers to have an internship. I just wanted to explore the field. Currently, I’m even interested in my core department, so I’m trying to do something in that. I have my CDC bagged in the field of analog electronics. So there I will be exploring more in my core department and then decide further.

How did you start with the project? What tips would you provide for people aspiring to go in the same field?

I already had taken up courses from Coursera on deep learning and deep CNN models. However, there wasn’t much content on Domain Adaptation. So, I had to resort to reading research papers and other reading materials. I had my code implemented in TensorFlow. I was very new only to the field, so what I did was is that I looked at some similar projects on the same topic. After finding many projects on the same, so I tried to look at the codes from GitHub and follow up as to what the person has approached. You have to learn the implementation of the codes. I didn’t know python earlier. So first I learned a bit of python for that. Thus you must learn from these data projects. There is plenty of reading material. You must first have some brief knowledge on the courses then you would know what the code implementation is doing.