Samsung SRIB interview experience (2018)
This was an online round which was held on Samsung’s own platform. The test duration was of 3 hours and one question was there. The constrains were like, you can do only 5 submissions, though you can compile and run the code any number of times. You can’t use any other library apart from the <stdio.h>/<iostream>. Your code must pass all the 10 test cases. The question was to find if the graph is bipartite or not. If yes, then print those vertices having color assigned as 1, given that the first source is of color 1.
Around 200 students appeared for this round, 44 were selected for the further rounds.
Next round was a F2F round. My interviewer was a good person and was more inclined towards logic and understanding of the problem. He went through my resume. I had done projects and internship in Computer vision and Machine Learning. He asked me the following questions:-
- What is the problem with (a+b)/2. The answer is related to the integer range overflow.
- What is priority queue?
- Implement Heap and write the code for Heapify
- Rotations of AVL tree. What are some other Data structures apart from AVL, RB and B+- Trees which are balanced? The answer is TREAP( A mixture of Tree and Heap)
- Given an array and queries having range of index as input, you have you find the no. of unique elements. I gave the basic solution. Then he asked about the range queries using segment trees.
- Detect cycle in an undirected graph without using dfs and stack. He wanted an answer related to disjoint sets
- Ho to pass a 2D array to a function?
- What are minimum spanning trees? How to find them?
Next few questions were related to machine learning and computer vision:
- What is gradient descent?
- Is it confirmed that we will reach the global minimum every single time using the gradient descent. If not how and why and what to do in this case?
- Difference between local minima and global minima.
- Which algorithm did I use in face recognition and what were the accuracy and why they were different in different algorithms?
- Methods used for feature selection
Other questions asked from my friends were like, non-recursive mutex, paging, mesh technology, OSI models.
This round was taken by someone else. Initially, He asked me where I belong to and how to reach there. Then he asked me following questions:
- What is PCA( Principal Component Analysis)? How does it works? How it is used for dimensionality reduction?
- What is ensembling? How you ensemble different models?
- What is random forest?
- He gave me one question to write the code, given a 2D grid with mines(2), walls(0), start as (1) and destination as (3), I have to find the shortest path from source to destination. You can move in all direction apart from the diagonal.
- He gave me question related to prediction of a target value.The question was like:
- You have been given a data of an Employee with EmpId, Salary, Working hours, Awards+ Bonus, Onsite(Y/N), Ratings and appraisal, Number of years with current Manager. You have to create a model which will predict whether an employee would resign or not. I used normalization, removal of outliers, replacing of Nulls with mean, median, mode. Binning the features into ranges. Then using the entropy and information gain to form a decision tree. I also gave him another approach of using a scoring model based on a function using weights assigned to each feature.
This was a kind of HR round, He just asked us basic questions like about the place we belong from, what is AI, what is Machine Learning in a layman term.
Finally I was selected for FTE + internship.