Adrish Dey

I’m an incoming CS PhD student at Boston University advised by Prof. Edward Chien. I recieved my Bachelor’s Degree in Computer Science and Engineering from Netaji Subhash Engineering College. Currently, I am based out of Kolkata. I’m also an independent researcher, interested in exploring complicated problems in geometric and topological data processing.

When I’m not working, you can find me flying ✈️ in Microsoft Flight Simulator.

Research Interests

Although my research interests spans a wide variety of topics, you can find me tweeting mostly about:

Geometric and Topological Deep Learning
Geometric Statistics
Geometry Processing
Topological Data Analysis
Optimal Transport
Generative Models
Implicit Deep Learning

You can find CV Here. If there’s something I can help you about, or if you want to discuss abount some cool research problems, feel free to drop me a 📨 @ adrishd.cse2017@nsec.ac.in. More social / academic URLs can be found below 👇

news

Apr 14, 2022	Excited to share I’m starting my PhD with Prof. Edward Chien at Boston University starting Fall 2022.
Oct 21, 2021	Excited to share, starting Novemeber, I am joining Weights and Biases as Machine Learning Engineer – Remote via Remote Inc. ✨
Mar 16, 2021	Excited to share I just accepted an offer for Summer Research Fellowship with Geometric Data Processing lab, at MIT CSAIL. 💫 🥳 🎉
Nov 1, 2020	Our work “TopoSampler: A Topology Constrained Noise Sampling for GANs” got accepted for spotlight presentation at Workshop on Topological Data Analysis and Beyond. Yay!
Dec 15, 2019	I’ll be joining rephrase.ai as a Research Intern working on GANs for the next 8 weeks!
May 9, 2019	I got into Google Summer of Code where I’ll be working with the TensorFlow Hub Team building pretrained modules. 🥳 UPDATE: The ESRGAN pretrained module can be found at: https://tfhub.dev/captain-pool/esrgan-tf2/1. And the codebase can be found at: https://github.com/captain-pool/GSOC

publications

NeurIPS TDA ’20

TopoSampler: A Topology Constrained Noise Sampling for GANs

Adrish Dey, Sayantan Das

2020

Abs PDF Poster Spotlight Talk Spotlight Slides

This work studies disconnected manifold learning in generative models in the light of point-set topology and persistent homology. Under this formalism, the topological similarity of latent space in generative models with the underlying manifold of data distribution facilitates better generalization. To achieve this, we introduce a topology-constrained noise sampler, responsible for mapping the samples from Gaussian spheres to a latent embedding space, which in turn is constrained to be topologically similar to the manifold underlying the data distribution. We study the effectiveness of this method in GANs for learning disconnected manifolds. This is ongoing research, with the current report containing preliminary empirical experiments.