Roy Urbach's Profile Picture

Roy Urbach

I enjoy cool models of cool things.

Particulary enthusiastic about Deep Learning and Artificial Intelligence, and exploring different learning rules.

MSc in Computational Neuroscience from the Weizmann Institute of Sciecne and BSc in Computer Science and Congitive and Brain Sciences from the Heberew University.

My Repositories

CLoSeR project image

CLoSeR

Code for the paper "Semantic stimuli representations emerge in brain-inspired self-supervising ensembles of neural networks" by Urbach and Schneidman (2025).

View Project
Flow Matching and Diffusion project image

Flow Matching and Diffusion from scratch

A toy repo and notebooks showing basic flow matching and diffusion models.

View Project
NeRF image

Neural Radiance Fields (NeRF) from scratch

A toy repo and notebooks showing basic NeRF

View Project
TreeGenerator project image

TreeGenerator

A fun python simulation of 2D trees inspired by L-system.

View Project
AlbumRank project image

AlbumRank

Saves your scores for songs and albums, and lets you view album scores. Uses Spotify API.

View Project
ClapSync project image

ClapSync

A model of a crowd synchronising its clapping in python using a local learning rule.

View Project
TreeGenerator project image

Minimal python code for ARC-AGI

Some programs to solve ARC-AGI puzzles for the NeurIPS 2025 Google Code Gold Championship competition.

View Project

Publications

Semantic stimuli representations emerge in brain-inspired self-supervising ensembles of neural networks.

Roy Urbach and Elad Schneidman (2025).

arXiv: 2510.14486 [q-bio.NC]. URL arxiv.org/abs/2510.14486

Functional Brain-to-Brain Transformation with No Shared Data.

Wasserman, Navve, Roman Beliy, Roy Urbach, and Michal Irani (2025).

arXiv: 2404.11143 [q-bio.NC]. URL arxiv.org/abs/2404.11143