"I have no special talents. I am only passionately curious. ~ Albert Einstein"
This is my old website kept for legacy reasons. For my current website visit here .
I am currently working as a Research Assistant at Indian Institute of Technology (IIT), Hyderabad. I am also a part-time student at the International Institute of Information Technology (IIIT), Hyderabad, where I am in the process of completing my Master’s thesis. I am looking for PhD opportunities in USA, Canada and Europe. If all goes according to plan, I hope to be enrolled as a full-time PhD student by next Fall-2018.
I am broadly interested in topics related to machine learning and more specifically into developing mathematical methods to understand it. My current research involves trying to incorporate “interpretability” or “explainability” in black box deep learning algorithms. For my Master’s thesis, I worked on probabilistic modelling of temporal relations in Monte Carlo simulations of molecular systems.
Take a look at my CV for more details about my research projects.
Aditya Chattopadhyay, Anirban Sarkar, Prantik Howlader, and Vineeth N Balasubramanian. “Grad-CAM++: Generalized Gradient-based Visual Explanations for Deep Convolutional Networks” Proceedings of IEEE Winter Conf. on Applications of Computer Vision (WACV2018).
Aditya Chattopadhyay, Siddharth Goyal, Koushik Kasavajhala, and U. Deva Priyakumar. “Role of Urea-Aromatic Stacking Interactions in Stabilizing the Aromatic Residues of the Protein in Urea Induced Denatured State.” Journal of the American Chemical Society (2017).
Aditya Chattopadhyay, Min Zheng, Mark P. Waller, and U. Deva Priyakumar. “Mining Macromolecule (Un)folding Paths Using Machine Learning and Graph Theory”. Poster presented at 5th Modeling of Chemical and Biological (Re)Activity (MCBR-5).
~ To gain more exposure to core Machine Learning research, I decided to join the Deep Learing lab of Prof. Vineeth N Balasubramanian, at the Indian Institute of Technology (IIT), Hyderabad.
~ Here, I worked on making these deep learning algorithms more interpretable. This is especially important in risk-critical tasks like Healthcare systems, self-driving cars etc. If we know why the machine does what it does, it would be much easier to have faith in it!
~ Our paper on explainability in CNNs using gradient information will appear in the IEEE Winter Conference on Applications of Computer Vision (WACV2018) in the algorithms track.
~ Learnt a lot about optimization methods from Prof. Vineeth, by attending his course on Convex Optimization in Machine Learning at IIT-H.
~ With a burgeoning interest in machine learning and probability, I chose probabilistic modelling of molecular system trajectories as the focus of my Master’s thesis.
~ Often molecular trajectories are analysed using Markov Chains, which reveal patterns on time-scales much larger than the length of a typical simulation time. However, such methods are limited by a necessity of explicit integration of Newton’s equations of motion to generate the trajectory.
~ More efficient sampling techniques exist (variants of Monte-Carlo methods) but these methods lack temporal ordering of trajectory snapshots, precluding the possibility of constructing a transition matrix, which is core to all Markov Chain based methods.
~ In my thesis, I try to address this problem by using Expectation Maximization on a hybrid gaussian-boltzman probability density function to learn these temporal dependencies from Monte Carlo simulation data.
~ Selected for a summer research internship at the Centre of Multiscale Theory and Computation, University of Munster in the lab of Prof. Mark Waller. Formulated a project involving machine learning to discern chemical structures from their infra-red spectra.
~ Worked in the lab of Prof. U Deva Priyakumar where we tried to understand the mysteries behind protein unfolding using in-silico methods. We discovered a novel mode of interaction through which urea molecules denature proteins which brings us a tiny step closer to solving the puzzle in the grand scheme of things! This work is published in JACS - 2017
~ Consistently in Dean’s List every semester since Aug ‘15, which is awarded to the top 15% students in a batch.
~ Qualified for ACM-ICPC Asia Regionals (Amritapuri).
~ Developed a SQL engine to select, insert, delete and join records in a relational database. Implemented external merge-sort to quickly sort through the records. sorting of records using external-merge sort.
~ Developed a web-app to simultaneously edit or view documents on a cloud server without having to download the files. Implemented a REST-API.
~ Inspired from the Japanese Anime, Beyblade, developed a 3D spinning top game with multiplayer facility in Unity.
~ Embarked on a five year long journey for a dual degree programme at the International Institute of Information Technology (IIIT), Hyderabad [B.Tech in Computer Science & MS in Computational Natural Sciences].
Drop me an email, if you are interested in knowing more!
aditya[dot]c[at]research[dot]iiit[dot]ac[dot]in, adityababiblue[at]gmail[dot]com