top of page

I am an Assistant Professor at the ECE Department of the Technion

 

My research is in Machine Learning and Optimization. I am focused on the design and analysis of efficient algorithms for a wide class of Machine Learning and decision making scenarios.

Short bio:

I did my post-doc at the Institute for Machine Learning at ETHZ working with Prof. Andreas Krause.  Previously, I did my PhD at the IE&M Department of the Technion, working under the guidance of Prof. Elad HazanBefore that, I completed my master's at the EE Department of the Technion under the guidance of Prof. Nahum Shimkin.

Office: Fishbach, 459     Contact: kfirylevy@technion.ac.il

papers

Publications

SLowcalSGD : Slow Query Points Improve Local-SGD for Stochastic Convex Optimization

    Tehila Dahan, Kfir Y. Levy                                                 

     To appear in NeurIPS, 2024.

Weight for Robustness: A Comprehensive Approach towards Optimal Fault-Tolerant Asynchronous ML Training

    Tehila Dahan, Kfir Y. Levy                                                 

     To appear in NeurIPS, 2024.

Fault Tolerant ML: Efficient Meta-Aggregation and Synchronous Training

    Tehila Dahan, Kfir Y. Levy                                                 

     In ICML, 2024[arXiv]

Dynamic Byzantine-Robust Learning: Adapting to Switching Byzantine Workers

    Ron Dorfman, Naseem Amin Yehya, Kfir Y. Levy                                                     

     In ICML, 2024[arXiv]

Bring Your Own (Non-Robust) Algorithm to Solve Robust MDPs by Estimating The Worst Kernel

    Uri Gadot, Kaixin Wang, Navdeep Kumar, Kfir Y. Levy, Shie Mannor                               

     In ICML, 2024[arXiv]

Private and Federated Stochastic Convex Optimization: Efficient Strategies for Centralized Systems

    Roie ReshefKfir Y. Levy

    In ICML, 2024

 

A Study of First-Order Methods with a Deterministic Relative-Error Gradient Oracle

    Nadav Hallak, Kfir Y. Levy            

    In ICML, 2024

Efficient Value Iteration for s-rectangular Robust Markov Decision Processes

    Navdeep Kumar, Kaixin Wang, Kfir Yehuda Levy, Shie Mannor        

    In ICML, 2024[arXiv]

Learning the Uncertainty Set in Robust Markov Decision Process

    Navdeep Kumar, Kaixin Wang, Uri Gadot, Kfir Y. Levy, Shie Mannor     

    In ICLR Tiny Papers Track, 2024[pdf] 

Policy Gradient with Tree Search (PGTS) in Reinforcement Learning Evades Local Maxima

    Navdeep Kumar, Priyank Agrawal, Kfir Y. Levy, Shie Mannor         

    In ICLR Tiny Papers Track, 2024[pdf]

Towards Faster Global Convergence of Robust Policy Gradient Methods

    Navdeep Kumar, Ilnura Usmanova, Kfir Yehuda Levy, Shie Mannor           

    In ICLR Tiny Papers Track, 2024[pdf]

Policy Gradient for Reinforcement Learning with General Utilities

    Navdeep Kumar, Kaixin Wang, Utkarsh Pratiush, Kfir Y. Levy, Shie Mannor

     In ICLR Tiny Papers Track, 2024[pdf]

Solving Non-rectangular Reward-robust MDPs via Frequency Regularization             

    Uri Gadot, Esther Derman, Navdeep Kumar, Maxence Mohamed Elfatihi, Kfir Levy, Shie Mannor

     In AAAI, 2024.  [arXiv]

Policy Gradient for Rectangular Robust Markov Decision Processes              

     Navdeep Kumar, Esther Derman, Matthieu Geist, Kfir Y. Levy, Shie Mannor

     In NeurIPS, 2023[pdf]

Meta-learning Adversarial Bandit Algorithms

    Mikhail Khodak, Ilya Osadchiy, Keegan Harris, Maria-Florina Balcan, Kfir Y. Levy, Ron Meir, Steven Wu

    In NeurIPS, 2023[arXiv]

Dropcompute: Simple and more Robust Distributed Synchronous Training via Compute Variance Reduction

    Niv Giladi, Shahar Gottlieb, Moran Shkolnik, Asaf Karnieli, Ron Banner, Elad Hoffer, Kfir Y. Levy, Daniel Soudry

     In NeurIPS, 2023[arXiv]

No-Regret Dynamics in the Fenchel Game: A Unified Framework

for Algorithmic Convex Optimization              

     Jun-Kun Wang, Jacob Abernethy and Kfir Y. Levy

     In Mathematical Programming, 2023[arXiv]

DocoFL : Downlink Compression for cross-device

Federated Learning

    Ron Dorfman, Shay Vargaftik, Yaniv Ben-Itzhak, and Kfir Y. Levy

     In ICML 2023. [arXiv]

Robust linear regression for general feature distribution

    Tom Norman, Nir Weinberger and Kfir Y. Levy

     In AISTATS 2023. [pdf][arXiv]

Explainable artificial intelligence (xai) techniques for energy and power systems: Review, challenges and opportunities.                    

     Ram Machlev, Leena Heistrene, Michael Perl, Kfir Y. Levy, Juri Belikov, Shie Mannor and Yoash Levron

     In Energy and AI 2022[pdf

Adapting to Mixing Time in Stochastic Optimization with Markovian Data

    Ron Dorfman, and Kfir Y. Levy

     In ICML 2022. [pdf][arXiv],  Spotlight Presentation

UNDERGRAD: A Universal Black-Box Optimization Method with Almost Dimension-Free Convergence Rate Guarantees

    Kimon Antonakopoulos, Dong Quan Vu, Volkan Cevher, Kfir Y. Levy, and Panayotis Mertikopoulos

     In ICML 2022. [pdf][arXiv],  Spotlight Presentation 

High probability bounds for a class of nonconvex algorithms with adagrad stepsize

    Kfir Y. Levy, Ali Kavis and Volkan Cevher

     In ICLR 2022[pdf][arXiv]

Faster Neural Network Training with Approximate Tensor Operations        

     Menachem Adelman, Kfir Y. Levy, Ido Hakimi and Mark Silberstein

     In NeurIPS 2021[pdf]

STORM+: Fully Adaptive SGD with Momentum for Non-convex Optimization   

    Kfir Y. Levy, Ali Kavis and Volkan Cevher

     In NeurIPS 2021[pdf][arXiv]

Learning Measuring Explainability and Trustworthiness of Power Quality

Disturbances Classifiers Using XAI - Explainable Artificial Intelligence      

    Ram Machlev, Michael Perl, Juri Belikov, Kfir Y. Levy and Yoash Levron

     In IEEE Transactions on Industrial Informatics 2021[pdf

LAGA: Lagged AllReduce with Gradient Accumulation for Minimal Idle Time

     Ido Hakimi, Rotem Aviv,  Kfir Y. Levy and Assaf Schuster 

     In ICDM 2021. [pdf]

Learning under Delayed Feedback: Implicitly Adapting to Gradient Delays     

    Rotem Aviv, Ido Hakimi, Assaf Schuster and Kfir Y. Levy

     In ICML 2021. [pdf[arXiv

Fast Projection onto Convex Smooth Constraints

    Ilnura Usmanuva, Kfir Y. Levy, Maryam Kamgarpour and Andreas Krause

     In ICML 2021. [pdf

Multi-Player Bandits: The Adversarial Case

    Pragnya Alatur, Kfir Y. Levy and Andreas Krause

     In JMLR 2020. [arXiv

Adaptive Sampling for Stochastic Risk-averse Learning 

    Sebastian Curi, Kfir Y. Levy, Stefanie Jegelka and Andreas Krause.

    In NeurIPS 2020. [pdf] 

Online Convex Optimization In the Random Order Model

    Dan Garber, Gal Korcia and Kfir Y. Levy 

    In ICML 2020.  [pdf]

UniXGrad: A Universal, Adaptive Algorithm with Optimal Guarantees for Constrained Optimization 

     Ali Kavis, Kfir Y. Levy, Francis Bach, and Volkan Cevher

    In NeurIPS 2019. [arXiv

Evaluating GANs via Duality

    Paulina Grnarova, Kfir Y. Levy, Aurelien Lucchi, Nathanael Perraudin,

     Ian Goodfellow, Thomas Hofmann and Andreas Krause

     In NeurIPS 2019. [arXiv

Online Variance Reduction with Mixtures

    Zalán BorsosSebastian Curi, Kfir Y. Levy, and Andreas Krause

     In ICML 2019. [pdf] 

A Universal Algorithm for Variational Inequalities Adaptive to Smoothness and Noise

    Francis Bach and Kfir Y. Levy

     In COLT 2019. [arXiv

Unsupervised Imitation Learning

    Sebastian Curi, Kfir Y. Levy, and Andreas Krause

     In CDC 2019. [arXiv] 

Projection Free Online Learning over Smooth Sets

    Kfir Y. Levy and Andreas Krause

     In AISTATS 2019. [pdf]

Online Adaptive Methods, Universality and Acceleration 

    Kfir Y. Levy, Alp Yurtsever, and Volkan Cevher

     In NeurIPS 2018. [pdf[arXiv

Online Variance Reduction for Stochastic Optimization

    Zalán BorsosAndreas Krause, and Kfir Y. Levy

     In COLT 2018. [pdf] [arXiv] [Code]

Faster Rates for Convex-Concave Games

    Jacob AbernethyKevin A. LaiKfir Y. Levy, and Jun-Kun Wang

     In COLT 2018. [pdf] [arXiv] 

An Online Learning Approach to Generative Adversarial Networks

    Paulina GrnarovaKfir Y. LevyAurelien LucchiThomas Hofmann,  and Andreas Krause

     In ICLR 2018. [pdf][arXiv

Online to Offline Conversions, Universality and Adaptive Minibatch Sizes

    Kfir Y. Levy

     In NIPS 2017. [pdf][arXiv

Continuous DR-submodular Maximization: Structure and Algorithms

    An BianKfir Y. LevyAndreas Krause, and Joachim M. Buhmann

     In NIPS 2017. [pdf[arXiv

k*-Nearest Neighbors: From Global to Local

    Oren Anava and Kfir Y. Levy

     In NIPS 2016. [pdf[arXiv] [Code]

On Graduated Optimization for Stochastic Non-Convex Problems

    Elad Hazan, Kfir Y. Levy, and Shai Shalev-Shwartz

     In ICML 2016. [pdf[arXiv][Code]

Faster Evasion of Saddle Points.

    Kfir Y. Levy

     Preprint 2016.[arXiv

Beyond Convexity: Stochastic Quasi-Convex Optimization

     Elad Hazan, Kfir Y. Levy, and Shai Shalev-Shwartz

     In NIPS 2015. [pdf[arXiv]

Fast Rates for Exp-concave Empirical Risk Minimization

     Tomer Koren and Kfir Y. Levy

     In NIPS 2015. [pdf]

Bandit Convex Optimization: Towards Tight Bounds

     Elad Hazan and Kfir Y. Levy

     In NIPS 2014. [pdf] [Full Version]

Logistic Regression: Tight Bounds for Stochastic and Online Optimization

     Elad Hazan, Tomer Koren and Kfir Y. Levy

     In COLT 2014. [pdf] [arXiv]

Unified Inter and Intra Options Learning Using Policy Gradient Methods

     Kfir Y. Levy and Nahum Shimkin

     In EWRL 2011. [pdf]

 

    

bottom of page