Ethan Fetaya

Ethan Fetaya is a machine-learning researcher whose work spans computer vision, relational inference and structured learning. He studied mathematics at the Hebrew University under Alex Lubotzky and pursued doctoral research at Weizmann under Shimon Ullman before becoming a postdoctoral fellow at the University of Toronto. His research includes human-pose and obstacle-detection networks, Neural Relational Inference, and a 2020 ICML Outstanding Paper on symmetric sets. Later, he coauthored 2025 work on conformal prediction with noisy labels and 2026 work on adversarial attacks in weight-space classifiers. As of October 2026, he is a Senior Lecturer at Bar-Ilan University's Alexander Kofkin Faculty of Engineering.

Now at: Bar-Ilan University

Timeline

2011

A master's thesis on homological codes

He studies mathematics at the Hebrew University under Alex Lubotzky, and his 2011 master's thesis, Homological Error Correcting Codes and Systolic Geometry, proves a square-root bound for homological codes from two-dimensional surfaces.
2016

Human pose by consensus

In 2016, with Ita Lifshitz and Shimon Ullman, he coauthors Human Pose Estimation using Deep Consensus Voting. A convolutional net lets each image location vote for keypoint positions, then combines the votes to identify a pose configuration.
2017–2018

Obstacle detection for autonomous driving

At the 2017 ICCV Workshops, he coauthors Real-Time category-based and general obstacle detection for autonomous driving. The work uses a unified deep convolutional network to combine category-based and general obstacle detection at real-time speed; IEEE publishes it in 2018.
2018

Neural relational inference in Toronto

After doctoral research in computer vision and machine learning in the Weizmann Institute's Department of Computer Science and Applied Mathematics under Shimon Ullman, he works as a postdoctoral fellow in the University of Toronto's machine-learning group with Rich Zemel and Max Welling. In 2018, he coauthors Neural Relational Inference for Interacting Systems, an unsupervised model that learns interactions and dynamics from observational data.
2020

An ICML Outstanding Paper

In 2020, On Learning Sets of Symmetric Elements earns an ICML Outstanding Paper Award. Coauthored with Haggai Maron, Or Litany and Gal Chechik, it introduces Deep Sets for Symmetric Elements layers that approximate invariant and equivariant functions.
2025

Noisy-label conformal prediction

In 2025, with Coby Penso and Jacob Goldberger, he coauthors Conformal Prediction of Classifiers with Many Classes based on Noisy Labels. Their Noise-Aware Conformal Prediction method estimates a noise-free threshold from validation data with noisy labels and provides a finite-sample coverage guarantee under uniform noise.
2026

Adversarial attacks in weight space

In 2026, he coauthors Adversarial Attacks in Weight-Space Classifiers, published in Transactions on Machine Learning Research.
As of October 2026

Senior Lecturer at Bar-Ilan

He is a Senior Lecturer at Bar-Ilan University's Alexander Kofkin Faculty of Engineering.