Yedid Hoshen Associate Professor, School of Computer Science and Engineering, the Hebrew University of Jerusalem Email: yedid.hoshen_at_mail.huji.ac.il

About Me

I am an Associate Professor at the School of Computer Science and Engineering, the Hebrew University of Jerusalem. I am also a visiting faculty rsearcher at Google.

Before joining HUJI, I was a Research Scientist at Facebook AI Research in New York And Tel Aviv. My current research interests include anomaly detection, learning disentangled representations and generative models. I completed my PhD at the Hebrew University of Jerusalem, supervised by Prof. Shmuel Peleg. Previously I studied Physics at the University of Oxford, traded options (volatility arbitrage) for a prop desk at KBC-FP, designed mission control algorithms for the Israeli Space Program and worked at Google for two summers (Google [x] and Google Research NYC).

Publications

ObjectDrop: Bootstrapping Counterfactuals for Photorealistic Object Removal and Insertion D. Winter, M. Cohen, S. Fruchter, Y. Pritch, A. Rav-Acha, Y. Hoshen ECCV 2024 project page

From Zero to Hero: Cold-Start Anomaly Detection T. Reiss, Y. Hoshen Findings of the ACL 2024 paper   |   code

Recovering the Pre-Fine-Tuning Weights of Generative Models E. Horwitz, J. Kahana, Y. Hoshen ICML 2024 project page   |   code

Disentanglement of single-cell data with biolord Z. Piran, N. Cohen, Y. Hoshen, M Nitzan Nature Biotechnology 2024 paper   |   code

Anomaly Detection and Biomarkers Localization in Retinal Images L. Tiosano, R. Abutbul, R. Lender, Y. Shwartz, I. Chowers, Y. Hoshen, J. Levy Journal of Clinical Medicine 2024 paper

Detecting anomalous proteins using deep representations T. Michael-Pitschaze, N. Cohen, D. Ofer, Y. Hoshen, M. Linial NAR Genomics and Bioinformatics 2024 paper   |   code

Unsupervised Word Segmentation Using Temporal Gradient Pseudo-Labels T.S. Fuchs, Y. Hoshen ICASSP 2023 paper   |   code

Red PANDA: Disambiguating Anomaly Detection by Removing Nuisance Factors N. Cohen, J Kahana, Y. Hoshen ICLR 2023 paper

Mean-shifted contrastive loss for anomaly detection T. Reiss, Y. Hoshen AAAI 2023 paper   |   code

You Say Factorization Machine, I Say Neural Network-It's All in the Activation C. Almagor, Y. Hoshen RecSys 2022 paper   |   code

Unsupervised Word Segmentation using K Nearest Neighbors T.S. Fuchs Y. Hoshen, J. Keshet Interspeech 2022 paper   |   code

A Contrastive Objective for Learning Disentangled Representations J. Kahana Y. Hoshen ECCV 2022 paper   |   code

Anomaly Detection Requires Better Representations T. Reiss, N. Cohen, E. Horwitz, R. Abutbul Y. Hoshen ECCV 2022 Workshops paper

The inductive bias of in-context learning: Rethinking pretraining example design Y. Levine, N. Wies, D. Jannai, D. Navon Y. Hoshen A. Shashua ICLR 2022 (Spotlight Presentation) paper

An Image is Worth More Than a Thousand Words: Towards Disentanglement in the Wild A. Gabbay, N. Cohen Y. Hoshen NeurIPS 2021 project page

Scaling-up Disentanglement for Image Translation A. Gabbay Y. Hoshen ICCV 2021 project page

Deep Single Image Manipulation Y. Vinker*, E. Horwitz*, N. Zabari Y. Hoshen ICCV 2021 (Oral Presentation) project page   |   code

Reconstruction-Based Membership Inference Attacks are Easier on Difficult Problems A. Shafran, S. Peleg Y. Hoshen ICCV 2021 paper

PANDA -- Adapting Pretrained Features for Anomaly Detection T. Reiss*, N. Cohen*, L. Bergman Y. Hoshen CVPR 2021 paper   |   code

Crypto-Oriented Neural Architecture Design A. Shafran, G. Segev, S. Peleg Y. Hoshen ICASSP 2021 paper

Classification-Based Anomaly Detection for General Data L. Bergman Y. Hoshen ICLR 2020 paper   |   code

Demystifying Inter-Class Disentanglement A. Gabbay Y. Hoshen ICLR 2020 paper   |   code

Neural Separation of Observed and Unobserved Distributions T. Halperin, A. Ephrat Y. Hoshen ICML 2019 paper   |   code

Non-Adversarial Image Synthesis with Generative Latent Nearest Neighbors Y. Hoshen, K. Li and J. Malik CVPR 2019 paper   |   code

Towards Unsupervised Single-Channel Blind Source Separation using Adversarial Pair Unmix-and-Remix Y. Hoshen ICASSP 2019 paper

Non-Adversarial Mapping with VAEs Y. Hoshen NIPS 2018 paper

Non-Adversarial Unsupervised Word Translation Y. Hoshen, L. Wolf EMNLP 2018 arXiv   |   code

NAM: Non-Adversarial Unsupervised Domain Mapping Y. Hoshen, L. Wolf ECCV 2018 arXiv   |   ICLR extended abstract   |   code

Unsupervised Correlation Analysis Y. Hoshen L. Wolf CVPR 2018 arXiv

Identifying Analogies Across Domains Y. Hoshen L. Wolf ICLR 2018 paper

VAIN: Attentional Multi-agent Predictive Modeling Y. Hoshen NIPS 2017 arXiv

An Egocentric Look at Video Photographer Identity Y. Hoshen S. Peleg CVPR 2016 arXiv   |   popular press

Visual Learning of Arithmetic Operations Y. Hoshen S. Peleg AAAI 2016 arXiv   |   code

Live Video Synopsis for Multiple Cameras Y. Hoshen S. Peleg ICIP 2015 paper

Speech Acoustic Modeling From Raw Multichannel Waveforms Y. Hoshen R.J. Weiss K.W. Wilson ICASSP 2015 paper

The Information in Temporal Histograms Y. Hoshen S. Peleg WACV 2015 paper

Wisdom of the Crowd in Egocentric Video Curation Y. Hoshen G. Ben-Artzi S. Peleg CVPR 2014 - Workshop paper

Efficient Representation of Distributions for Background Subtraction Y. Hoshen Y. Poleg C. Arora S. Peleg AVSS 2013 paper

Resume

My resume can be found here.