IISc MTech (Research) Interview Experience (CD-CS)

20 May 2019
Gate score - 762, AIR- 388, Category - General

CDS is the Department of Computational and Data sciences in IISc. Research admissions in CDS dept are taken in 2 streams- CD-CS and CD-CP. For more details, refer IISc-CDS We had to fill a Google doc well before the interview dates, in which we had to choose a stream from these 2, and assign a preference for the labs in which we want to work. The labs in the streams are given in the admissions brochure in the link.

There was a written test for 30 mins with 5 questions based on which candidates were shortlisted for interviews. The questions were easy, based on probability, combinatorics, programming, complex nos. Gate level preparation is enough to get selected.

This year there was requirement for 4 labs in CD-CS-
VAL Lab, VCL Lab, DREAM Lab, MARS Lab. Subjects to be prepared for each lab will be given in Admissions brochure. There were 2 interview panels, one for DREAM lab (Say panel A) and another was for VAL, VCL Lab (Say panel B). I had put VAL as 1st pref, VCL as 2nd and DREAM Lab as 3rd pref. MARS Lab was accepting only PhD candidates. There were around 10-12 candidates for VAL, VCL Lab interviews.

Well Panel B was taking around 40-50 mins for each candidate, and panel A was going comparitively fast. So soon all candidates for Panel A were done, and 3 people from Panel B were sent to wait near Panel A, and I was the 3rd. But here was the twist, Panel A was asking questions on programming, ALgorithms, Data strcutures, whereas B was asking on Linear Algebra, Probability. So it would be a loss for those students who had opted for VAL lab, and not prepared for Algo, DS much. The 1st candidate was from Electrical background, so he was again sent back to Panel B. The 2nd candidate, who was a good friend of mine, was inerviewed by Panel A, and he was asked about programming and Algo. Since I also had an EC background, I was more comfortable with Maths, so I decided to request the panel to send me to Panel B. Interview at Panel A:

Interviewer: (Here Prof. Yogesh Simmhan was asking me questions. Reads out my entire profile. Asked few general questions about my BTech in EC, then Gate in CS, and about my work Ex, but not much)
Me: Sir, just a request, I had prepared Maths topics, so I would be more comfortable if I am asked those questions more.

Interviewer: Ohh, so you didn’t had Algorithms in your Btech?
Me: No Sir, I have prepared it for Gate

Interviewer: So atleast you know a bit of DS and Algo since gate was in CS.
Me: Yes sir. (They were considerate enough and asked me basic questions)

Interviewer: Do you know what is Post-order traversal
Me: I explained

Interviewer: What is Pre-order and In-order
Me: I explained

Interviewer: Draw a random Binary tree, a bit complex one and write down it’s Post-order traversal
Me: I drew a tree and wrote the Post-order

Interviewer : Write In-order and Pre-order also
Me: Did that with ease

Interviewer : Can we get back the tree if we have just the Pre-Order and Post-Order or any other 2 combinations?
Me: No Sir, Preorder and Postorder are not enough. We must have In-Order to get the tree back.

Interviewer: Give a logic so as to get the tree from In-Order and Post-Order.
Me: I explained this, it was the standard Gate question and they were impressed.

Interviewer: Okay, so do you want to work in Dream lab, or you have given it as 3rd preference, just because you had to give something as 3rd pref?
Me: No Sir, I’m not much interested in Dream lab, I just gave it as 3rd pref because I had to give something

Interviewer: Fair enough, no issues. So we are done with your basic interview, and for more advanced questions wait near Panel B.
Me: I heaved a sigh of relief.

(It was already around 6.00 pm, and there were 4 candidates left including me at Panel B, and my turn was the last. Finally at around 7.30 pm I was called. My entire interview was taken by Prof. Venkatesh Babu.)

Interview at Panel B: Interviewer: (Read my profile, so you were interviewed once and were waiting since long?)
Me: Yes Sir, actually it was my choice.

Interviewer: So what have you prepared?
Me: Sir, Probability and Linear Algebra.

Interviewer: Let X be a Random variable, and X denote the number of tosses required till first head appears. Then what will be the expected value of this Random variable?
Me: I knew this. Sir, X denotes a Geometric Random variable. Then I drew the table of possible X values with their probability and did entire derivation for the Expected value of this distribution.

Interviewer: He was impressed with my approach. Then he asked, I have coins, (any no of coins), and I want to perform the experiment of throwing a die. First he checked whether I knew die throwing experiment properly. Then if this experiment must be done using just coins, how we will do it?
Me: I was trying with 2 coins, but they were giving just 4 possible outcomes. So he said to increase the no of coins.They were giving lot of hints, and finally

Interviewer: You can use any number of coins
Me: Then I took 4 coins, and I thought we can map the number of heads with the face value of a die. But this way all the faces of die would not be equi-probable. So counting number of heads was not making a die throw experiment.

Interviewer: They were giving many hints. He then asked if x is no of coins, then what are no of outcomes possible?
Me: 2^x sir.

Interviewer: Then this 2^x is equal to some natural number m times 6.
Me: I wrote 2^x=6m and was trying to use this hint. It meant 2^{x-1}=3m. But no such m and x exists. So eventually, this die throw experiment cannot be done with equal probability for all faces.

Interviewer: Let A be a matrix such that A = UU^T, B = U^TU. Now if \Lambda is eigen value of A, you have to prove that \Lambda is the eigen value of B. And under what conditions this exist.
Me: I didn’t knew this from before. I started with A = \Lambda X and we need to prove B = \Lambda Y. With little bit help from the profs, I managed to prove it.

Interviewer: He extended the previous question. If U is a column matrix, what will be the rank of A and B
Me: I took an example and got it right.

Interviewer: What will be the eigen values for A and B
Me: Told that.

Interviewer: Do know anything about Machine learning or have you studied something about the lab.
Me: (After my past experiences, I had thought that I won’t tell that I know about Machine Learning as I was not much confident) No sir, I just know a bit application wise, but I have not done any course

Interviewer: Something basic, like do you know about classification and regression?
Me: I explained it in simple terms

Interviewer: He then asked me to give an example for Linear Regression
Me: I drew the common housing and prices fitting graph, and explained.

Interviewer: He asked about what the terms in graph denote
Me: Explained the graph

Interviewer: Then what are the parameters here and how do we fix them?
Me: I told weights and biases are parameters and we take a cost function and minimize it to get the optimal parameter values.

Interviewer: He then asked me to write the cost function for Linear Regression
Me: I wrote that.

Interviewer: How do we minimize this?
Me: Sir by taking the derivatives partially.

Interviewer: Take it and show
Me: I wrote whatever I could get and he too helped a lot in deriving these parameter values.

(That’s it. Then they asked me if I am applying for MTech or PhD. They were like we have very limites seats and first preference we are giving for PhD candidates. I said I want to go for MTech.)

Results: On 24th May, interview for all candidates at CDS were done and provisional list was released. In CD-CS they had given a list of 16 candidates, and my application number was on 14th number!! I was very happy when I got the mail from IISc. This I feel was indeed a dream come true… :)

Suggestions:
Professors are chilled, they will give you lot of hints, and lot of time to think, only if you are progressing in explaining them. So even if you don’t know any answer or proof beforehand, just start with the basic definitions, they will help you ahead. Infact they even told me during some proof that don’t try this way, you won’t get the right results.
And yeah make sure you are interviewed by the panel who is asking subjects related to your expertise.
For Probability, I referred these MIT Lectures And for Linear Algebra, Gilbert Strang’s videos

Thanks for reading guys, and if you liked my post, don’t forget to share among your friends for whom this might be helpful. Cheers!


© 2018. All rights reserved.

Powered by Hydejack v8.4.0