From Grundy
Jump to: navigation, search

A series of informal talks conducted by senior students. The speaker describes his/her internship or research experience. The key theme is Computer Science or programming research work. This article lists down the various Reflections events that have taken place in the past.

List of Reflections

  • 21st August, 2016 - Pi Calculus for Security Protocols, Kushal Babel - Applied Pi calculus is a formalism for specifying behavior of security protocols and other concurrent processes. Why is this formalism important and how has it helped us in the past?
  • 28th August, 2016 - Forwarding Schemes in Switched Networks with Probabilistic Faults, Shubham Goel
  • 25th September, 2016 - Approximation Algorithms, Sanat Anand - Sanat Anand spoke on two problems in computer science namely - the traveling repairman problem and the online job scheduling problem. He described NP hardness and various approximation algorithms for the audience. This is related to his internship in Summer '16.
  • 6th January, 2017 - Online Algorithms for Swarm Robotics, Siddhant Garg - Siddhant worked on several online algorithmic problems and even attended a course on Online Algorithms during his stay at TU Braunschweig. His primary work was designing and analyzing algorithms for a problem of swarm robotics. The problem was about a swarm of robots exploring an unknown area trying to minimize the time of exploration and the energy of the robots. Another problem which he worked on was on efficiently packing square shapes in a constrained area.
  • 12th January, 2017 - SCION Computer Architecture, ETH Zurich, Shantanu Thakoor, Sivaprasad Sudhir - They spoke about SCION, a future internet architecture proposed by researchers at ETH Zurich. It is a clean slate architecture that aims to provide scalability, security, and control as an integral part of the network.
  • 19th January 2017 - What Personality Do Brands Have?, Pradyot Prakash - Did you know that brands have a character? While Red bull portrays ruggedness, Louis Vuitton represents sophistication and caters to the elite. Through this talk, he tried to draw a parallel between the brands and the personality traits evoked by them. The discussion touched on the work he did during his internship at Adobe Research. This was a first of a kind of work connecting marketing and areas in Computer Science. He talked about how textual articles published by companies to extract these personality traits and how machine learning models can be used to check for consistency in the tone carried by the articles.
  • 2nd February 2017 - Cloud Computing and Big Data, Chandra Maloo - Cloud Computing and Big Data are terms that have been trending for quite a while now. By harnessing the capability of cloud for big data analytics one can produce astounding results. In this talk, he discussed about a couple of frameworks available for big data analysis by reflecting upon the work he did last summer in his internship at Samsung Electronics. The project was about analyzing usage data of Samsung Smart Home digital appliances to gain insights into consumer behaviour and the performance of Samsung appliances.
  • 12th February 2017 - Resource Analysis for the Leon System (EPFL), Sumith Kulal - Resource analysis is a current active research area wherein one tries to reason about the resource consumption of a program/system. This information, of say space and time complexity, is naturally important for both theory and practice alike. For example, prevention of side channel attack by not leaking any information gained from the physical implementation of a cryptosystem. Sumith will be presenting the latest work done on this for the Leon verification system ( which is published in the proceedings of POPL 2017. This includes new approach for specifying and verifying resource utilization of higher-order functional programs that use lazy evaluation and memoization. Resource consumption of lazy evaluation is particularly hard to reason about. This work was done as a part of Sumith's internship at EPFL.
  • 3rd March 2017 - Optimal Setting of Hyperparameters, Ritwick Chaudhary - Machine Learning algorithms are rarely parameter free and often it seems appealing to develop machine learning algorithms with fewer of them. But some complex models involve a lot of hyperparameters and it is extremely hard to find out an optimal setting for the paramters. Typically Random Search or Grid Search or the programmer’s experience helps in setting the hyper parameters for a learning model but can’t be applied if the number of hyperparameters is a lot. For tuning of such hyperparameters, we could consider the problem as a problem involving optimization of a black-box function where the function to be modelled is the learning model’s performance on the training data and the domain is the space of hyperparameters. A package called Spearmint which has been developed by the University of Toronto and Harvard University was very helpful in trying out various approaches and to run the algorithms in parallel.
  • 22nd September 2017 - Recurrent Learning, Ritwick Chaudhary - Ritwick Chaudhry will be talking about his project at Adobe Research. The project was basically on Machine Learning in e-learning platforms. Here is a detailed Description: To kick off this year's Reflections, Ritwick Chaudhry will be talking about the project he worked on during his summer internship at Adobe Research. This project was a great success and is being patented and submitted as a research paper. His project was on personalizing learners’ experience on e-learning platforms. A futuristic and important part of e-learning systems like MOOCs is being intelligent and being able to gauge the students skills and learning styles. Assessments are also a major component of the e-learning experience which help to track learning progress in different concepts and the project was based on trying to model students' skill from their previous assessments and their propensity to take a hint rather than attempting.

See also