Avatar

Amartya Sanyal

DPhil (PhD) Student in Computer Science

University of Oxford

Biography

I am Amartya Sanyal, currently a D.Phil(PhD) student at the Department of Computer Science at the University of Oxford working with Prof. Varun Kanade and Prof. Phil H.S. Torr. I completed by Bachelor Of Technology in Computer Science And Engineering from the Indian Institute of Technology, Kanpur in 2017.

I am interested in understanding why Deep Neural Networks, in whatever form, perform as good as they do. My research revolves around understanding mathematical properties regarding the structures of these networks and designing and analysing optimization algorithms that can exploit their properties. In the process, I also hope to come up with some practical applications using these findings. I am also interested in Optimization and Learning Theory in general.

Interests

  • Optimization
  • Generalization
  • Regularization

Education

  • PhD in Computer Science, 2017-2021

    University of Oxford

  • BTech in Computer Science, 2013-2017

Recent Posts

Robustness via Deep Low-Rank Representations

In this paper ( Full Paper here), we investigate the relation of the intrinsic dimension of the representation space of deep networks …

NonConvexity of neural networks

I don't remember where I heard this arguement but it was a pretty interesting one and I couldn't remember the entire reasoning and so …

Strong Convexity and Strong Smoothness

I am going to , inspired by the course on optimization that I am doing this semester, talk a bit about strong convexity and strong …

Montreal

Montreal from Mont Royal As I said before, during the Summers of 2016, I went to Montreal to work as a Research Intern in …

Visits

Internship

I have had the opportunity to visit a few places and carry out my research there. Here is a list of the places I have visited.

Work

  • Computer Vision And Pattern Recognition Unit, ISI Kolkata - A hybrid architecture for face recignition in real life scenario.
  • Machine Learning Team, Amazon.com, Bangalore Worked on a competitive pricing model.
  • Montreal Institute of Learning Algorithms Worked on a novel Multiscale Language Model with a learned dictionary.
  • Twitter Cortex, New York Worked there during the summer of 2017.
  • Krakow, Poland Visisted Krakow to attend BESC 2017 where I delivered a 30 min oral presentation.
  • Buenos Aired Attended the Machine Learning Summer School in Buenos Aires.
  • Stockholm, Sweden Visited Stockholm in 2018 to attend COLT and ICML. (1 Poster + 2 Workshop Posters)
  • Long Beach, LA, US Visited Long Beach in 2019 to attend ICML. (1 Workshop Poster)
  • Genoa, Italy Spent a couple of Months at LCSL in Genoa, Italy with Lorenzo Rosasco.
  • Addis Adaba Visiting Addis Adaba this summer to present our paper on Stable Rank Normalization in ICML 2020.

Not Work

  • Visited almost the entirety of India.
  • Hiked for a day in parts of Quebec around Val David, Val Morin.
  • Short trips to Paris, Edinburgh, Brussels, and Copenhagen.

Contact