Tom Le Paine

Senior Research Scientist at DeepMind.
Working on large-scale deep reinforcement learning, and using demonstrations to make it more efficient and effective.
Previously: Applying deep learning to computer vision — speeding it up, and making it work with less labeled data. PhD student in the Image Formation and Processing group at UIUC, advised by Professor Thomas Huang.
Google scholar



Google AI
Summer 2015

Google Brain
Summer 2014

Spring 2014

Summer 2013

2008 - 2009


AlphaStar: Mastering the Real-Time Strategy Game StarCraft II

The AlphaStar Team

blog match video making of video

One-Shot High-Fidelity Imitation: Training Large-Scale Deep Nets with RL

Tom Le Paine, Sergio Gomez Colmenarejo, Ziyu Wang, Scott Reed, Yusuf Aytar, Tobias Pfaff, Matt W. Hoffman, Gabriel Barth-Maron, Serkan Cabi, David Budden, Nando de Freitas

NeurIPS 2018 Deep Reinforcement Learning Workshop (Oral)

Large-Scale Visual Speech Recognition

Brendan Shillingford, Yannis Assael, Matthew W. Hoffman, Tom Le Paine, Cían Hughes, Utsav Prabhu, Hank Liao, Hasim Sak, Kanishka Rao, Lorrayne Bennett, Marie Mulville, Ben Coppin, Ben Laurie, Andrew Senior, Nando de Freitas


Playing Hard Exploration Games by Watching YouTube

Yusuf Aytar, Tobias Pfaff, David Budden, Tom Le Paine, Ziyu Wang, Nando de Freitas

NeurIPS 2018

Few-shot Autoregressive Density Estimation: Towards Learning to Learn Distributions

Scott Reed, Yutian Chen, Tom Le Paine, Aäron van den Oord, S. M. Ali Eslami, Danilo Rezende, Oriol Vinyals, Nando de Freitas

ICLR 2018

Fast Generation for Convolutional Autoregressive Models

Tom Le Paine, Prajit Ramachandran, Pooya Khorrami, Mohammad Babaeizadeh, Shiyu Chang, Yang Zhang, Mark A. Hasegawa-Johnson, Roy H. Campbell, Thomas S. Huang

ICLR 2017 (Workshop)

How Deep Neural Networks Can Improve Emotion Recognition on Video Data

Pooya Khorrami, Tom Le Paine, name, name, Thomas S. Huang

ICIP 2016

Seq-NMS for Video Object Detection

Tom Le Paine, Wei Han, Pooya Khorrami, Prajit Ramachandran, Mohammad Babaeizadeh, Honghui Shi, Jianan Li, Shuicheng Yan, Thomas S. Huang

ICCV 2015 (Workshop)

Do Deep Neural Networks Learn Facial Action Units When Doing Expression Recognition?

Pooya Khorrami, Tom Le Paine, Thomas S. Huang

ICCV 2015 (Workshop)

An Analysis of Unsupervised Pre-training in light of recent advances

Tom Le Paine, Pooya Khorrami, Wei Han, Thomas S. Huang

ICLR 2015 (Workshop)
PDF Code (Github)

Visual Media: History and Perspectives

Thomas S. Huang, Vuong Le, Tom Le Paine, Pooya Khorrami, Usman Tariq

IEEE Multimedia 2014

GPU asynchronous stochastic gradient descent to speed up neural network training

Tom Le Paine, Hailin Jin, Jianchao Yang, Zhe Lin, Thomas S. Huang

ICLR 2014 (Workshop)

Deep learning for face recognition

Tom Le Paine, Thapanapong Rukkanchanunt

Technical report

Simultaneous dynamic and functional MRI scanning (SimulScan) of natural swallows

Tom Le Paine, Charles Conway, Georgia Malandraki, Bradley Sutton

Magnetic Resonance in Medicine

Examination of susceptibility effects on functional and dynamic magnetic resonance imaging

Tom Le Paine

UIUC Master's thesis

Optimized preload leakage-correction methods to improve the diagnostic accuracy of dynamic susceptibility-weighted contrast-enhanced perfusion MR imaging in posttreatment gliomas

LS Hu, LC Baxter, DSPinnaduwage, TL Paine, JP Karis, BG Feuerstein, KM Schmainda, AC Dueck, J Debbins, KA Smith, P Nakaji, JM Eschbacher, SW Coons, JE Heiserman

American Journal of Neuroradiology



A python micro-framework for training neural networks. Built on top of Theano. Written with the help of my colleague Pooya Khorrami.

from anna import layers
input = layers.Input()
conv1 = layers.Conv2DLayer(


Spring 2013
Teaching assistant for CS 543: Computer Vision