Or Sharir, Ph.D.
I am a Postdoctoral Scholar at the Computing + Mathematical Sciences (CMS) Department at the California Institute of Technology (Caltech), advised by Prof. Anima Anandkumar and Prof. Garnet Chan. My research focuses on investigating and further developing the theoretical underpinnings for deep learning methods, with some surprising connections to Quantum Physics.
I obtained my Ph.D. in Computer Science from the Hebrew University of Jerusalem under the supervison of Prof. Amnon Shashua. I also have a B.Sc. in Physics, Mathematics and Computer Science from the Hebrew University. (Detailed CV).
If you are looking for my old class notes, you can find them here (website and notes are in Hebrew).
- or sharir.org
- Annenberg 225.
- Began my Postdoctoral appointment at Caltech.—Sep 24th, 2021.
- Spotlight by Google's TPU Research Cloud program.—Sep 20th, 2021.
- I am the recipeint of the 2020 Dan David Prize Young Researchers Scholarship.—May 5th, 2020.
- We released a paper developing specialized deep autoregressive models for the efficient simulation of quantum systems.—February 11th, 2019.
- Our paper analysing the quantum entanglment supported by deep learning architectures was accepted to the prestige journal, Physical Review Letters.—January 4th, 2019.
- Two of our papers, “On the Expressive Power of Overlapping Architectures of Deep Learning” and “Benefits of Depth for Long-Term Memory of Recurrent Networks”, were accepted for presentation at ICLR 2018, the former on the conference track and the latter on the workshop track.—January 30th, 2018.
- Our paper “Sum-Product-Quotient Networks” was accepted to AISTATS 2018.—December 22nd, 2017.
- We released a paper theoretically analyzing the benefits of depth to deep RNNs.—October 25th, 2017.
- We released a paper proposing an extension to Sum-Product Networks that exponentially boosts their expressive power.—October 12th, 2017.
- We released a paper on merging our tensorial analysis of ConvNets with Sum-Product Networks.—October 13th, 2016.
- Our paper “On the Expressive Power of Deep Learning: A Tensor Analysis” was accepted to COLT 2016.—April 26th, 2016.
- Our paper “Deep Simnets” was accepted to CVPR 2016.—Feburary 29th, 2016.
- FlowKet — A Python framework for variational Monte-Carlo simulations of many-body quantum systems on top of Tensorflow.
- Github-MathJax (Chrome Extension) — a chrome extension for rendering LaTeX equations in Github repositories. Very useful for documenting research projects.
- reprochart — A usefuly python script for easily reproducible charts.
- Generative ConvACs (Experiments) — scripts for reproducing our experiments on Generative ConvACs.
- SimNets (Caffe Fork) — our fork of Caffe with the original implementation of the SimNets architecture.
Neural tensor contractions and the expressive power of deep neural quantum states. NeurIPS 2021 @ ML4Science Workshop.
Originally released on arXiv on March 19th, 2021.
Limits to Depth Efficiencies of Self-Attention. NeurIPS 2020.
Originally released on arXiv on June 22nd, 2020.
SenseBERT: Driving Some Sense into BERT. ACL 2020.
Originally released on arXiv on August 15th, 2019.
Deep Autoregressive Models for the Efficient Variational Simulation of Many-body Quantum Systems. Physical Review Letters.
Originally released on arXiv on Feburary 11th, 2019.
Quantum Entanglement in Deep Learning Architectures. Physical Review Letters.
Originally released on arXiv on March 26th, 2018.
Benefits of Depth for Long-Term Memory of Recurrent Networks. ICLR 2018 Workshop.
Originally released on arXiv on October 25th, 2017.
Sum-Product-Quotient Networks. AISTATS 2018.
Originally released on arXiv on October 12th, 2017.
On the Expressive Power of Overlapping Architectures of Deep Learning. ICLR 2018.
Originally released on arXiv on March 6th, 2017.
Tensorial Mixture Models. Preprint. (Source Code)
Originally released on arXiv on October 13th, 2016.
On the Expressive Power of Deep Learning: A Tensor Analysis. COLT 2016.
Originally released on arXiv on September 16th, 2015.
Deep SimNets. CVPR 2016. (Source Code)
Originally released on arXiv on June 9th, 2015.