Nikhil Mehta

I am a Ph.D. Candidate at Duke University. My research area is deep learning and I am advised by Professor Lawrence Carin.

Prior to joining Duke, I spent 18 wonderful months at IIT Kanpur as a research associate working with Professor Piyush Rai and Professor Gaurav Pandey. I hail from the capital city of India, Delhi, where I received my bachelor's degree from the Delhi Technological University.

Email / Google Scholar / Linkedin / Github

Internships

Research Intern, Facebook AI Research (FAIR) Accel

Jan - April 2022
Research Intern, Google Brain

May - August 2021
Research Intern, Google AutoLX

May - August 2020

Research

	Efficient Feature Transformations for Discriminative and Generative Incremental Learning V. Verma, K. Liang, N. Mehta, P. Rai, & L. Carin CVPR, 2021 A task-specific feature map transformation strategy for continual learning, which adds minimal parameters to the base architecture while outerperformaing previous methods in discriminative and generative tasks.
	WAFFLe: Weight Anonymized Factors for Federated Learning W. Hao, N. Mehta, K. Liang, P. Cheng, M. Khamy & L. Carin IEEE Access, Volume 10 A Bayesian nonparametric framework using shared rank-1 weight factors for federated learning.
	Continual Learning using a Bayesian Nonparametric Dictionary of Weight Factors N. Mehta, K. Liang, V. Verma, & L. Carin AISTATS, 2021 Continual Learning using Bayesian Nonparametrics and layer-wise dictionary of weight factors.
	Counterfactual Representation Learning with Balancing Weights S. Assaad, S. Zeng, C. Tao, S. Datta, N. Mehta, R. Henao, F. Li, & L. Carin AISTATS, 2021 Counterfactual Representation Learning with Balancing Weights
	Graph Representation Learning via Ladder Gamma VAE. A. Sarkar^, N. Mehta^, & P. Rai AAAI, 2020. Early version appeared at Graph Representation Learning Workshop (NeurIPS). A Gamma Ladder Variational Autoencoder for graph-structured data that brings together the interpretability in hierarchical, multilayer latent variable models and the strong representational power of graph encoders.
	Stochastic Blockmodels meet Graph Neural Networks N. Mehta, L. Carin & P. Rai ICML, 2019 A deep generative framework for overlapping community discovery and link prediction combining the interpretability of stochastic blockmodels, such as the latent feature relational model, with the modeling power of deep generative models.
	Estimation and Sampling of Unnormalized Statistical Models with Stein Score Matching N. Mehta^, Jiachang Liu^, Chenyang Tao & L. Carin, Workshop on Stein’s Method. ICML, 2019 We propose using Stein's Method to estimate the parameters of a distribution based on given samples. We also estimate the score function of the empirical distribution and propose a new generative model combining Stein score matching with the Langevin flow.
	Deep Topic Models for Multi-Label Learning R. Panda, A. Pensia, N. Mehta, M.Zhou & P. Rai AISTATS, 2019 We present a probabilistic framework for multi-label learning based on a deep generative model for the binary label vector associated with each observation.
	Survival Cluster Analysis P. Chapfuwa, C. Li, N. Mehta, L. Carin and R. Henao CHIL, 2019 A Bayesian nonparametric model for jointly inferring the time-to-event and the individualized risk-based cluster assignments.
	Robust and Fast 3D Scan Alignment using Mutual Information N. Mehta, J. McBride & G. Pandey ICRA, 2018 A mutual information based algorithm for the estimation of full 6-degree-of-freedom (DOF) rigid body transformation between two overlapping point clouds.

Forked from Jon Barron's page.