About Me


My name is Reuben. I am currently working at Common Sense Machines, a Boston-based AI company founded by MIT and DeepMind alumni. Previously, I completed my Ph.D. in the Human & Machine Learning Lab at NYU, funded by a Google Ph.D. Fellowship in Computational Neuroscience. With my thesis advisor, Brenden Lake, I developed a new computational framework to model the structure of human conceptual knowledge, emphasizing everyday concepts like animals, vehicles, and handwritten characters. By reconciling key modeling ingredients from deep neural networks and structured Bayesian models, my thesis helps provide a comprehensive account for the flexibility of human concepts. While at NYU, I completed an internship at Facebook AI Research (FAIR) where I worked directly under chief AI scientist Yann LeCun developing self-supervised learning algorithms.

Before NYU, I earned my Sc.B. in Applied Mathematics at Brown University, where I completed my thesis in computational vision with Thomas Serre and Stuart Geman. Following undergrad, I worked at Symantec's Center for Advanced Machine Learning in Mountain View, CA as a full-time research engineer for two years. There, I helped found a new paradigm for the company's threat detection technologies based on artificial intelligence, filing 8 U.S. patents and improving the detection of malicious software on over 100 million customer machines worldwide.

CV

Research


Publications

Compositional Diversity in Visual Concept Learning (Jan 2024)
Yanli Zhou, Reuben Feinman, Brenden M. Lake
Cognition, 244, 105711

Generative Neuro-Symbolic Models of Concept Learning (Sep 2023)
Reuben Feinman
Ph.D. thesis, New York University

Learning Task-General Representations with Generative Neuro-Symbolic Modeling (Jun 2020)
Reuben Feinman, Brenden M. Lake
International Conference on Learning Representations (ICLR 2021)
[Code]

Generating New Concepts with Hybrid Neuro-Symbolic Models (Mar 2020)
Reuben Feinman, Brenden M. Lake
In Proceedings of the 42nd Annual Conference of the Cognitive Science Society

A Linear Systems Theory of Normalizing Flows (Feb 2020)
Reuben Feinman, Nikhil Parthasarathy
arXiv preprint arXiv:1907.06496
[Code]

Learning a Smooth Kernel Regularizer for Convolutional Neural Networks (Mar 2019)
Reuben Feinman, Brenden M. Lake
In Proceedings of the 41st Annual Conference of the Cognitive Science Society
[Code]

Learning Inductive Biases with Simple Neural Networks (Feb 2018)
Reuben Feinman, Brenden M. Lake
In Proceedings of the 40th Annual Conference of the Cognitive Science Society
[Code] [Poster]

Detecting Adversarial Samples from Artifacts (Mar 2017)
Reuben Feinman, Ryan R. Curtin, Saurabh Shintre, Andrew B. Gardner
arXiv preprint arXiv:1703.00410
[Code]

Reports

Cleverhans v1.0.0: An Adversarial Machine Learning Library (Dec 2016)
Nicolas Papernot, Ian Goodfellow, Ryan Sheatsley, Reuben Feinman, Patrick McDaniel
arXiv preprint arXiv:1610.00768
[Code]

A Deep Belief Network Approach to Learning Depth from Optical Flow (May 2015)
Reuben Feinman
Sc.B. Honors thesis, Brown University

Select U.S. Patents

Providing Adversarial Perturbations to Media (Feb 2020)
Saurabh Shintre, Reuben Feinman
US No. 10,542,034

Systems and Methods for Trichotomous Malware Classification (Jul 2019)
Reuben Feinman, Javier Echauz, Andrew B. Gardner
US No. 10,366,233

Systems and Methods for Detecting Malware Based on Event Dependencies (May 2019)
Jugal Parikh, Reuben Feinman
US No. 10,282,546

Systems and Methods for Detecting Malware (Nov 2018)
Reuben Feinman, Jugal Parikh
US No. 10,133,865


Talks


Structure and Emergence in Human Concepts

NYU Neuroscience department presentation
[Slides]

Oct 20, 2020

Learning a Smooth Kernel Regularizer for Convolutional Neural Networks

NYU Computational Cognitive Science group meeting.
[Slides]

Feb 27, 2019

Learning Inductive Biases with Neural Networks

NYU CILVR group meeting.
[Slides]

Feb 26, 2018

Background


Positions


  • Sept 2023
    -
    Present

    Common Sense Machines

    Research Scientist

    Developing the next generation of 3D creation engines with AI.

  • Sept 2017
    -
    Aug 2023

    New York University

    Graduate Research Assistant

    Conducted my doctoral research at NYU's Center for Neural Science, advised by Brenden Lake.

  • May 2020
    -
    Sept 2020

    Facebook AI Research (FAIR)

    Research Intern

    Worked with chief AI scientist Yann LeCun investigating self-supervised learning algorithms for computer vision applications.

  • Jul 2015
    -
    Jun 2017

    Symantec Corporation

    Machine Learning Engineer

    Worked as a full-time research engineer at Symantec's Center for Advanced Machine Learning, a newly chartered R&D lab with a focus on state-of-the-art pattern recognition technologies.

  • Sept 2014
    -
    May 2015

    Brown University

    Undergraduate Research Assistant

    Conducted undergraduate research in the Serre Lab at Brown under supervision of Thomas Serre and Stuart Geman.


Education


  • Sept 2017
    -
    Present

    New York University

    Ph.D., Neural Science

    Current doctoral student.

  • Sept 2011
    -
    May 2015

    Brown University

    Sc.B., Applied Mathematics

    Graduated with Honors. GPA: 3.9/4.0.

Blog


A Purpose for Nondeterminism: Statistical Inference in the Brain

Randomness isn’t just a quantum byproduct—it serves a critical role in neural computation.

Mar 25, 2017

Evolution's Paramount Gift: The Ability to Learn

To solve intelligence, we must solve the learning dynamics of the brain.

July 8, 2016

20 Questions and Information Theory

Information Theory is arguably one of the most beautiful and profound set of ideas that our race has developed.

Mar 27, 2016

Increasing the Efficiency of Labor

A new paradigm for technological development has risen to prominence in the 21st century.

Dec 16, 2015

Why A.I. Makes Government Surveillance Exponentially More Powerful

News flash: no human monitor is necessary in order for surveillance programs to understand a person’s text messages in the truest sense of the word!

Oct 7, 2015

Knowledge Atoms and Representational Features

Recently, researchers have devised learning algorithms that enable machines to develop high-level knowledge structures.

Aug 31, 2015

DeepMind's Subtle Revolution

DeepMind's Atari agent learns to play video games in the same way that infant mammals learn to perform various functions.

Aug 17, 2015

The Power of Information

The process by which intelligent systems develop assigns an inherent value to data.

Aug 9, 2015

Contact


4 Washington Place, Room 809
New York, NY 10003

reuben.feinman@nyu.edu

+1 212 998 7780
(Center for Neural Science)