Get our free extension to see links to code for papers anywhere online!

Chrome logo Add to Chrome

Firefox logo Add to Firefox
Picture for Kannan Ramchandran

Kannan Ramchandran

Model Selection with Near Optimal Rates for Reinforcement Learning with General Model Classes


Jul 13, 2021
Avishek Ghosh, Sayak Ray Chowdhury, Kannan Ramchandran

* 24 pages 
  Access Paper or Ask Questions

Model Selection for Generic Contextual Bandits


Jul 07, 2021
Avishek Ghosh, Abishek Sankararaman, Kannan Ramchandran

Figure 1 for Model Selection for Generic Contextual Bandits
* 40 pages, 5 figures. arXiv admin note: text overlap with arXiv:2006.02612 
  Access Paper or Ask Questions

Collaborative Learning and Personalization in Multi-Agent Stochastic Linear Bandits


Jun 15, 2021
Avishek Ghosh, Abishek Sankararaman, Kannan Ramchandran

Figure 1 for Collaborative Learning and Personalization in Multi-Agent Stochastic Linear Bandits
Figure 2 for Collaborative Learning and Personalization in Multi-Agent Stochastic Linear Bandits
* 25 pages, 8 figures 
  Access Paper or Ask Questions

LocalNewton: Reducing Communication Bottleneck for Distributed Learning


May 16, 2021
Vipul Gupta, Avishek Ghosh, Michal Derezinski, Rajiv Khanna, Kannan Ramchandran, Michael Mahoney

Figure 1 for LocalNewton: Reducing Communication Bottleneck for Distributed Learning
Figure 2 for LocalNewton: Reducing Communication Bottleneck for Distributed Learning
Figure 3 for LocalNewton: Reducing Communication Bottleneck for Distributed Learning
Figure 4 for LocalNewton: Reducing Communication Bottleneck for Distributed Learning
* To be published in Uncertainty in Artificial Intelligence (UAI) 2021 
  Access Paper or Ask Questions

Escaping Saddle Points in Distributed Newton's Method with Communication efficiency and Byzantine Resilience


Mar 17, 2021
Avishek Ghosh, Raj Kumar Maity, Arya Mazumdar, Kannan Ramchandran

Figure 1 for Escaping Saddle Points in Distributed Newton's Method with Communication efficiency and Byzantine Resilience
Figure 2 for Escaping Saddle Points in Distributed Newton's Method with Communication efficiency and Byzantine Resilience
Figure 3 for Escaping Saddle Points in Distributed Newton's Method with Communication efficiency and Byzantine Resilience
Figure 4 for Escaping Saddle Points in Distributed Newton's Method with Communication efficiency and Byzantine Resilience

  Access Paper or Ask Questions

Provably Breaking the Quadratic Error Compounding Barrier in Imitation Learning, Optimally


Feb 25, 2021
Nived Rajaraman, Yanjun Han, Lin F. Yang, Kannan Ramchandran, Jiantao Jiao

Figure 1 for Provably Breaking the Quadratic Error Compounding Barrier in Imitation Learning, Optimally
Figure 2 for Provably Breaking the Quadratic Error Compounding Barrier in Imitation Learning, Optimally
* 30 pages, 2 figures 
  Access Paper or Ask Questions

BEAR: Sketching BFGS Algorithm for Ultra-High Dimensional Feature Selection in Sublinear Memory


Oct 26, 2020
Amirali Aghazadeh, Vipul Gupta, Alex DeWeese, O. Ozan Koyluoglu, Kannan Ramchandran

Figure 1 for BEAR: Sketching BFGS Algorithm for Ultra-High Dimensional Feature Selection in Sublinear Memory
Figure 2 for BEAR: Sketching BFGS Algorithm for Ultra-High Dimensional Feature Selection in Sublinear Memory
Figure 3 for BEAR: Sketching BFGS Algorithm for Ultra-High Dimensional Feature Selection in Sublinear Memory
Figure 4 for BEAR: Sketching BFGS Algorithm for Ultra-High Dimensional Feature Selection in Sublinear Memory

  Access Paper or Ask Questions

Fast Distributed Training of Deep Neural Networks: Dynamic Communication Thresholding for Model and Data Parallelism


Oct 18, 2020
Vipul Gupta, Dhruv Choudhary, Ping Tak Peter Tang, Xiaohan Wei, Xing Wang, Yuzhen Huang, Arun Kejariwal, Kannan Ramchandran, Michael W. Mahoney

Figure 1 for Fast Distributed Training of Deep Neural Networks: Dynamic Communication Thresholding for Model and Data Parallelism
Figure 2 for Fast Distributed Training of Deep Neural Networks: Dynamic Communication Thresholding for Model and Data Parallelism
Figure 3 for Fast Distributed Training of Deep Neural Networks: Dynamic Communication Thresholding for Model and Data Parallelism
Figure 4 for Fast Distributed Training of Deep Neural Networks: Dynamic Communication Thresholding for Model and Data Parallelism
* 17 pages, 8 figures 
  Access Paper or Ask Questions

FastSecAgg: Scalable Secure Aggregation for Privacy-Preserving Federated Learning


Sep 23, 2020
Swanand Kadhe, Nived Rajaraman, O. Ozan Koyluoglu, Kannan Ramchandran

Figure 1 for FastSecAgg: Scalable Secure Aggregation for Privacy-Preserving Federated Learning
Figure 2 for FastSecAgg: Scalable Secure Aggregation for Privacy-Preserving Federated Learning
Figure 3 for FastSecAgg: Scalable Secure Aggregation for Privacy-Preserving Federated Learning
Figure 4 for FastSecAgg: Scalable Secure Aggregation for Privacy-Preserving Federated Learning
* Shorter version accepted in ICML Workshop on Federated Learning, July 2020, and CCS Workshop on Privacy-Preserving Machine Learning in Practice, November 2020 
  Access Paper or Ask Questions

Boundary thickness and robustness in learning models


Jul 09, 2020
Yaoqing Yang, Rajiv Khanna, Yaodong Yu, Amir Gholami, Kurt Keutzer, Joseph E. Gonzalez, Kannan Ramchandran, Michael W. Mahoney

Figure 1 for Boundary thickness and robustness in learning models
Figure 2 for Boundary thickness and robustness in learning models
Figure 3 for Boundary thickness and robustness in learning models
Figure 4 for Boundary thickness and robustness in learning models

  Access Paper or Ask Questions

Problem-Complexity Adaptive Model Selection for Stochastic Linear Bandits


Jun 15, 2020
Avishek Ghosh, Abishek Sankararaman, Kannan Ramchandran

Figure 1 for Problem-Complexity Adaptive Model Selection for Stochastic Linear Bandits
* 24 pages, 8 figures 
  Access Paper or Ask Questions

An Efficient Framework for Clustered Federated Learning


Jun 07, 2020
Avishek Ghosh, Jichan Chung, Dong Yin, Kannan Ramchandran

Figure 1 for An Efficient Framework for Clustered Federated Learning
Figure 2 for An Efficient Framework for Clustered Federated Learning
Figure 3 for An Efficient Framework for Clustered Federated Learning
* 20 pages, 4 figures and 1 table 
  Access Paper or Ask Questions

Communication-Efficient Gradient Coding for Straggler Mitigation in Distributed Learning


May 14, 2020
Swanand Kadhe, O. Ozan Koyluoglu, Kannan Ramchandran

Figure 1 for Communication-Efficient Gradient Coding for Straggler Mitigation in Distributed Learning
Figure 2 for Communication-Efficient Gradient Coding for Straggler Mitigation in Distributed Learning
Figure 3 for Communication-Efficient Gradient Coding for Straggler Mitigation in Distributed Learning
* Shorter version accepted in 2020 IEEE International Symposium on Information Theory (ISIT) 
  Access Paper or Ask Questions

Alternating Minimization Converges Super-Linearly for Mixed Linear Regression


Apr 23, 2020
Avishek Ghosh, Kannan Ramchandran

Figure 1 for Alternating Minimization Converges Super-Linearly for Mixed Linear Regression
Figure 2 for Alternating Minimization Converges Super-Linearly for Mixed Linear Regression
Figure 3 for Alternating Minimization Converges Super-Linearly for Mixed Linear Regression
Figure 4 for Alternating Minimization Converges Super-Linearly for Mixed Linear Regression
* Accepted for publication at AISTATS, 2020 
  Access Paper or Ask Questions

Communication-Efficient and Byzantine-Robust Distributed Learning


Nov 21, 2019
Avishek Ghosh, Raj Kumar Maity, Swanand Kadhe, Arya Mazumdar, Kannan Ramchandran

Figure 1 for Communication-Efficient and Byzantine-Robust Distributed Learning
Figure 2 for Communication-Efficient and Byzantine-Robust Distributed Learning
Figure 3 for Communication-Efficient and Byzantine-Robust Distributed Learning
Figure 4 for Communication-Efficient and Byzantine-Robust Distributed Learning

  Access Paper or Ask Questions

Max-Affine Regression: Provable, Tractable, and Near-Optimal Statistical Estimation


Jun 21, 2019
Avishek Ghosh, Ashwin Pananjady, Adityanand Guntuboyina, Kannan Ramchandran

Figure 1 for Max-Affine Regression: Provable, Tractable, and Near-Optimal Statistical Estimation
Figure 2 for Max-Affine Regression: Provable, Tractable, and Near-Optimal Statistical Estimation
Figure 3 for Max-Affine Regression: Provable, Tractable, and Near-Optimal Statistical Estimation
Figure 4 for Max-Affine Regression: Provable, Tractable, and Near-Optimal Statistical Estimation
* The first two authors contributed equally to this work and are ordered alphabetically 
  Access Paper or Ask Questions

Robust Federated Learning in a Heterogeneous Environment


Jun 16, 2019
Avishek Ghosh, Justin Hong, Dong Yin, Kannan Ramchandran

Figure 1 for Robust Federated Learning in a Heterogeneous Environment
Figure 2 for Robust Federated Learning in a Heterogeneous Environment
* 30 pages, 4 figures 
  Access Paper or Ask Questions

Adversarially Trained Autoencoders for Parallel-Data-Free Voice Conversion


May 09, 2019
Orhan Ocal, Oguz H. Elibol, Gokce Keskin, Cory Stephenson, Anil Thomas, Kannan Ramchandran

Figure 1 for Adversarially Trained Autoencoders for Parallel-Data-Free Voice Conversion
Figure 2 for Adversarially Trained Autoencoders for Parallel-Data-Free Voice Conversion
Figure 3 for Adversarially Trained Autoencoders for Parallel-Data-Free Voice Conversion
Figure 4 for Adversarially Trained Autoencoders for Parallel-Data-Free Voice Conversion

  Access Paper or Ask Questions

Gradient Coding Based on Block Designs for Mitigating Adversarial Stragglers


Apr 30, 2019
Swanand Kadhe, O. Ozan Koyluoglu, Kannan Ramchandran

Figure 1 for Gradient Coding Based on Block Designs for Mitigating Adversarial Stragglers
Figure 2 for Gradient Coding Based on Block Designs for Mitigating Adversarial Stragglers
Figure 3 for Gradient Coding Based on Block Designs for Mitigating Adversarial Stragglers
Figure 4 for Gradient Coding Based on Block Designs for Mitigating Adversarial Stragglers
* Shorter version accepted in 2019 IEEE International Symposium on Information Theory (ISIT) 
  Access Paper or Ask Questions

OverSketched Newton: Fast Convex Optimization for Serverless Systems


Mar 21, 2019
Vipul Gupta, Swanand Kadhe, Thomas Courtade, Michael W. Mahoney, Kannan Ramchandran

Figure 1 for OverSketched Newton: Fast Convex Optimization for Serverless Systems
Figure 2 for OverSketched Newton: Fast Convex Optimization for Serverless Systems
Figure 3 for OverSketched Newton: Fast Convex Optimization for Serverless Systems
Figure 4 for OverSketched Newton: Fast Convex Optimization for Serverless Systems
* 27 pages, 11 figures 
  Access Paper or Ask Questions

Cross-Entropy Loss and Low-Rank Features Have Responsibility for Adversarial Examples


Jan 24, 2019
Kamil Nar, Orhan Ocal, S. Shankar Sastry, Kannan Ramchandran

Figure 1 for Cross-Entropy Loss and Low-Rank Features Have Responsibility for Adversarial Examples
Figure 2 for Cross-Entropy Loss and Low-Rank Features Have Responsibility for Adversarial Examples
Figure 3 for Cross-Entropy Loss and Low-Rank Features Have Responsibility for Adversarial Examples
Figure 4 for Cross-Entropy Loss and Low-Rank Features Have Responsibility for Adversarial Examples

  Access Paper or Ask Questions

Frank-Wolfe Algorithm for Exemplar Selection


Nov 06, 2018
Gary Cheng, Armin Askari, Laurent El Ghaoui, Kannan Ramchandran

Figure 1 for Frank-Wolfe Algorithm for Exemplar Selection
Figure 2 for Frank-Wolfe Algorithm for Exemplar Selection
Figure 3 for Frank-Wolfe Algorithm for Exemplar Selection
Figure 4 for Frank-Wolfe Algorithm for Exemplar Selection

  Access Paper or Ask Questions

Rademacher Complexity for Adversarially Robust Generalization


Oct 29, 2018
Dong Yin, Kannan Ramchandran, Peter Bartlett

Figure 1 for Rademacher Complexity for Adversarially Robust Generalization
Figure 2 for Rademacher Complexity for Adversarially Robust Generalization
Figure 3 for Rademacher Complexity for Adversarially Robust Generalization

  Access Paper or Ask Questions

Defending Against Saddle Point Attack in Byzantine-Robust Distributed Learning


Sep 14, 2018
Dong Yin, Yudong Chen, Kannan Ramchandran, Peter Bartlett

Figure 1 for Defending Against Saddle Point Attack in Byzantine-Robust Distributed Learning
Figure 2 for Defending Against Saddle Point Attack in Byzantine-Robust Distributed Learning
Figure 3 for Defending Against Saddle Point Attack in Byzantine-Robust Distributed Learning

  Access Paper or Ask Questions

Online Scoring with Delayed Information: A Convex Optimization Viewpoint


Jul 09, 2018
Avishek Ghosh, Kannan Ramchandran

Figure 1 for Online Scoring with Delayed Information: A Convex Optimization Viewpoint
Figure 2 for Online Scoring with Delayed Information: A Convex Optimization Viewpoint
Figure 3 for Online Scoring with Delayed Information: A Convex Optimization Viewpoint
Figure 4 for Online Scoring with Delayed Information: A Convex Optimization Viewpoint
* 8 pages, 4 figures 
  Access Paper or Ask Questions

Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates


Mar 05, 2018
Dong Yin, Yudong Chen, Kannan Ramchandran, Peter Bartlett

Figure 1 for Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates
Figure 2 for Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates
Figure 3 for Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates
Figure 4 for Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates

  Access Paper or Ask Questions

Speeding Up Distributed Machine Learning Using Codes


Jan 29, 2018
Kangwook Lee, Maximilian Lam, Ramtin Pedarsani, Dimitris Papailiopoulos, Kannan Ramchandran

Figure 1 for Speeding Up Distributed Machine Learning Using Codes
Figure 2 for Speeding Up Distributed Machine Learning Using Codes
Figure 3 for Speeding Up Distributed Machine Learning Using Codes
Figure 4 for Speeding Up Distributed Machine Learning Using Codes
* This work is published in IEEE Transactions on Information Theory and presented in part at the NIPS 2015 Workshop on Machine Learning Systems and the IEEE ISIT 2016 
  Access Paper or Ask Questions

Gradient Diversity: a Key Ingredient for Scalable Distributed Learning


Jan 07, 2018
Dong Yin, Ashwin Pananjady, Max Lam, Dimitris Papailiopoulos, Kannan Ramchandran, Peter Bartlett

Figure 1 for Gradient Diversity: a Key Ingredient for Scalable Distributed Learning
Figure 2 for Gradient Diversity: a Key Ingredient for Scalable Distributed Learning
Figure 3 for Gradient Diversity: a Key Ingredient for Scalable Distributed Learning
Figure 4 for Gradient Diversity: a Key Ingredient for Scalable Distributed Learning

  Access Paper or Ask Questions

Approximate Ranking from Pairwise Comparisons


Jan 04, 2018
Reinhard Heckel, Max Simchowitz, Kannan Ramchandran, Martin J. Wainwright

Figure 1 for Approximate Ranking from Pairwise Comparisons
Figure 2 for Approximate Ranking from Pairwise Comparisons
Figure 3 for Approximate Ranking from Pairwise Comparisons
* AISTATS 2017 
  Access Paper or Ask Questions