Itay Safran's Homepage

About Me

I'm a Post-Doctoral Research Associate at Purdue University, hosted by Paul Valiant. Previously, I was a Postdoctoral Research Fellow at Princeton University, where I was hosted by Jason D. Lee. Prior to coming to Princeton, I completed my M.Sc. and Ph.D. at the Mathematics and Computer Science Department at the Weizmann Institute of Science, where I was greatly fortunate to be advised by Ohad Shamir. Prior to that, I did a dual B.Sc. in Computer Science and Mathematics at Ben-Gurion University of the Negev.

My research focuses on theoretical machine learning, with emphasis on deep learning theory. In particular, I am interested in the intersection of deep learning theory with related fields, such as approximation theory, optimization and theoretical computer science.

In October 2024, I will be joining the Computer Science Department at Ben-Gurion University of the Negev as an Assistant Professor.

Email: safran.itay at gmail dot com.

Office: Lawson 1208.

Publications

Preprints

Depth Separations in Neural Networks: Separating the Dimension from the Accuracy

Itay Safran, Daniel Reichman and Paul Valiant

arXiv technical report

Journal and Conference Papers

How Many Neurons Does it Take to Approximate the Maximum?

Itay Safran, Daniel Reichman and Paul Valiant

35th ACM-SIAM Symposium on Discrete Algorithms (SODA 2024)

On the Effective Number of Linear Regions in Shallow Univariate ReLU Networks: Convergence Guarantees and Implicit Bias

Itay Safran*, Gal Vardi* and Jason D. Lee

* Equal contribution

36th Conference on Neural Information Processing Systems (NeurIPS 2022)

Optimization-Based Separations for Neural Networks

Itay Safran and Jason D. Lee

35th Conference on Learning Theory (COLT 2022)

Random Shuffling Beats SGD Only After Many Epochs on Ill-Conditioned Problems

Itay Safran and Ohad Shamir

35th Conference on Neural Information Processing Systems (NeurIPS 2021 - Spotlight Presentation)

The Effects of Mild Over-parameterization on the Optimization Landscape of Shallow ReLU Neural Networks

Itay Safran*, Gilad Yehudai* and Ohad Shamir

* Equal contribution

34th Conference on Learning Theory (COLT 2021)

How Good is SGD with Random Shuffling?

Itay Safran and Ohad Shamir

33rd Conference on Learning Theory (COLT 2020)

Depth Separations in Neural Networks: What is Actually Being Separated?

Itay Safran, Ronen Eldan and Ohad Shamir

32nd Conference on Learning Theory (COLT 2019)

Spurious Local Minima are Common in Two-Layer ReLU Neural Networks

Itay Safran and Ohad Shamir

35th International Conference on Machine Learning (ICML 2018 - Long Presentation)

Depth-Width Tradeoffs in Approximating Natural Functions with Neural Networks

Itay Safran and Ohad Shamir

34th International Conference on Machine Learning (ICML 2017)

On the Quality of the Initial Basin in Overspecified Neural Networks

Itay Safran and Ohad Shamir

33rd International Conference on Machine Learning (ICML 2016)

Other Publications

A Simple Explanation for the Existence of Adversarial Examples with Small Hamming Distance

Adi Shamir, Itay Safran, Eyal Ronen and Orr Dunkelman

arXiv Technical Report

Page updated

Google Sites

Report abuse