From Enormous Structured Models to On-device Federated Learning: Robustness, Heterogeneity and Optimization.
Krishna Pillutla
PhD Dissertation (2022).
PDF   Slides

Federated Learning with Partial Model Personalization.
Krishna Pillutla, Kshitiz Malik, Abdelrahman Mohamed, Michael Rabbat, Maziar Sanjabi, Lin Xiao.
ICML 2022.
PDF   Code   Slides (ICML Spotlight)   Poster  

Robust Aggregation for Federated Learning.
Krishna Pillutla, Sham Kakade, Zaid Harchaoui.
IEEE Transactions on Signal Processing (2022).
PDF   Code (TensorFlow)   Code (PyTorch)   Talk video

MAUVE: Measuring the Gap Between Neural Text and Human Text using Divergence Frontiers.
Krishna Pillutla, Swabha Swayamdipta, Rowan Zellers, John Thickstun, Sean Welleck, Yejin Choi, Zaid Harchaoui.
NeurIPS 2021. Outstanding Paper Award (Top 6 out of 9000 submissions).
PDF   Pip-package   Code   Poster   Press

Divergence Frontiers for Generative Models: Sample Complexity, Quantization Level, and Frontier Integral.
Lang Liu, Krishna Pillutla, Sean Welleck, Sewoong Oh, Yejin Choi, Zaid Harchaoui.
NeurIPS 2021.
PDF   Code

LLC: Accurate, Multi-purpose Learnt Low-dimensional Binary Codes.
Aditya Kusupati, Matthew Wallingford, Vivek Ramanujan, Raghav Somani, Jae Sung Park, Krishna Pillutla, Prateek Jain, Sham Kakade, Ali Farhadi.
NeurIPS 2021.

Federated Learning with Heterogeneous Devices: A Superquantile Optimization Approach.
Krishna Pillutla*, Yassine Laguel*, Jérôme Malick, Zaid Harchaoui.

Superquantiles at Work : Machine Learning Applications and Efficient (Sub)gradient Computation.
Yassine Laguel, Krishna Pillutla, Jérôme Malick, Zaid Harchaoui.
Set-Valued and Variational Analysis (2021).
PDF     Publisher’s Page

A Superquantile Approach to Federated Learning with Heterogeneous Devices.
Yassine Laguel*, Krishna Pillutla*, Jérôme Malick, Zaid Harchaoui.
CISS 2021.
PDF PDF-arXiv(old)   Code

A Smoother Way to Train Structured Prediction Models.
Krishna Pillutla, Vincent Roulet, Sham Kakade, Zaid Harchaoui.
NeurIPS 2018.
PDF-long   PDF-short   Code   Documentation
Poster   Blog post   Video summary

A Markov Chain Theory Approach to Characterizing the Minimax Optimality of Stochastic Gradient Descent (for Least Squares).
Prateek Jain, Sham M. Kakade, Rahul Kidambi, Praneeth Netrapalli, Venkata Krishna Pillutla, Aaron Sidford.
FSTTCS 2017.

Data Driven Resource Allocation for Distributed Learning.
Travis Dick, Mu Li, Venkata Krishna Pillutla, Colin White, Maria-Florina Balcan, Alex Smola.
PDF-long   PDF-short  

On Skewed Multi-dimensional Distributions: the FusionRP Model, Algorithms, and Discoveries.
Venkata Krishna Pillutla, Zhanpeng Fang, Christos Faloutsos, Danai Koutra, Jie Tang.
SDM 2016.

Master’s Thesis: Data Driven Resource Allocation For Distributed Machine Learning.
Thesis Committee: Nina Balcan, Alex Smola, Christos Faloutsos
PDF Slides