Tom Schaul Dr.

Research scientist

Google - DeepMind

Location
London, United Kingdom
Discipline
Computer and Information Science
Interests
Recurrent Neural Networks, Reinforcement Learning, Artificial Curiosity, Optimization, AI in Games.
more... less...

About

I am doing a postdoc in machine learning at NYU in Yann LeCun's lab. Just recently I completed my Ph.D. at IDSIA/TU Munich, supervised by Jürgen Schmidhuber. I grew up in Luxembourg, and completed my studies in computer engineering at the EPFL in Switzerland, at Columbia University and at the University of Waterloo. I work on (modular) reinforcement learning, black-box optimization, (deep and/or recurrent) neural networks, artificial curiosity and related topics.
more... less...

Publications

Unit Tests for Stochastic Optimization

Schaul T, Antonoglou I, Silver D

International Conference on Learning Representations (2014)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Natural Evolution Strategies

Wierstra D, Schaul T, Glasmachers T, Sun Y, Peters J, Schmidhuber J

Journal of Machine Learning Research (2014)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

No More Pesky Learning Rates

Schaul T, Zhang S, LeCun Y

International Conference on Machine Learning (ICML) (2013)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Adaptive learning rates and parallelization for stochastic, sparse, non-smooth gradients

Schaul T, LeCun Y

International Conference on Learning Representations (2013)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

A Video Game Description Language for Model-based or Interactive Learning

Schaul T

Proceedings of the IEEE Symposium on Computational Intelligence and Games (CIG) (2013)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Better Generalization with Forecasts

Schaul T, Ring M

International Joint Conference on Artificial Intelligence (IJCAI) (2013)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

A Linear Time Natural Evolution Strategy for Non-Separable Functions

Sun Y, Gomez F, Schaul T, Schmidhuber J

Proceedings of the Genetic and Evolutionary Computation Conference (2013)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Adaptive learning rates and parallelization for stochastic, sparse, non-smooth gradients

Schaul T, LeCun Y

International Conference on Learning Representations (2013)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Optimization with Surrogate Models

Schaul T

Numerical Methods for Metamaterial Design (2013)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Towards a Video Game Description Language

Ebner M, Levine J, Lucas S, Schaul T, Thompson T, Togelius J

Dagstuhl Follow-up (2013)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

General Video Game Playing

Levine J, Bates Congdon C, Bída M, Ebner M, Kendall G, Lucas S et al.

Dagstuhl Follow-up (2013)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Scalable Neural Networks for Board Games

Schaul T, Schmidhuber J

International Conference on Artificial Neural Networks (ICANN) (2009)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

No More Pesky Learning Rates

Schaul T, Zhang S, LeCun Y

NIPS Workshop on Optimization for Machine Learning (OPT2012) (2012)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

The Organization of Behavior into Temporal and Spatial Neighborhoods

Ring M, Schaul T

Second Joint IEEE International Conference on Developmental Learning and Epigenetic Robotics (2012)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

No More Pesky Learning Rates

Schaul T, Zhang S, LeCun Y

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Investigating the Impact of Adaptation Sampling in Natural Evolution Strategies on Black-box Optimization Testbeds

Schaul T

Black-box Optimization Benchmarking Workshop, Genetic and Evolutionary Computation Conference (2012)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Comparing Natural Evolution Strategies to BIPOP-CMA-ES on Noiseless and Noisy Black-box Optimization Testbeds

Schaul T

Black-box Optimization Benchmarking Workshop, Genetic and Evolutionary Computation Conference (2012)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Benchmarking Exponential Natural Evolution Strategies on the Noiseless and Noisy Black-box Optimization Testbeds

Schaul T

Black-box Optimization Benchmarking Workshop, Genetic and Evolutionary Computation Conference (2012)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Benchmarking Separable Natural Evolution Strategies on the Noiseless and Noisy Black-box Optimization Testbeds

Schaul T

Black-box Optimization Benchmarking Workshop, Genetic and Evolutionary Computation Conference (2012)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Benchmarking Natural Evolution Strategies with Adaptation Sampling on the Noiseless and Noisy Black-box Optimization Testbeds

Schaul T

Black-box Optimization Benchmarking Workshop, Genetic and Evolutionary Computation Conference (2012)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Decoupling the Data Geometry from the Parameter Geometry for Stochastic Gradients

Schaul T, Zhang S, LeCun Y

Snowbird Learning Workshop (2012)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Natural Evolution Strategies Converge on Sphere Functions

Schaul T

Genetic and Evolutionary Computation Conference (GECCO) (2012)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Measuring Intelligence through Games

Schaul T, Togelius J, Schmidhuber J

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Natural Evolution Strategies

Wierstra D, Schaul T, Glasmachers T, Sun Y, Schmidhuber J

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

The Two-Dimensional Organization of Behavior

Ring M, Schaul T, Schmidhuber J

AAAI Workshop on Lifelong Learning from Sensorimotor Experience (2011)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

A Linear Time Natural Evolution Strategy for Non-Separable Functions

Sun Y, Gomez F, Schaul T, Schmidhuber J

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

The Two-Dimensional Organization of Behavior

Ring M, Schaul T, Schmidhuber J

First Joint IEEE International Conference on Developmental Learning and Epigenetic Robotics (2011)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Studies in Continuous Black-box Optimization

Schaul T

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Coherence Progress: A Measure of Interestingness Based on Fixed Compressors

Schaul T, Pape L, Glasmachers T, Graziano V, Schmidhuber J

Fourth Conference on Artificial General Intelligence (AGI) (2011)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

High Dimensions and Heavy Tails for Natural Evolution Strategies

Schaul T, Glasmachers T, Schmidhuber J

Genetic and Evolutionary Computation Conference (GECCO) (2011)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Artificial Curiosity for Autonomous Space Exploration

Graziano V, Glasmachers T, Schaul T, Pape L, Cuccu G, Leitner J et al.

Acta Futura (2011)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Q-error as a Selection Mechanism in Modular Reinforcement-Learning Systems.

Ring M, Schaul T

International Joint Conference on Artificial Intelligence (IJCAI) (2011)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Multi-Dimensional Deep Memory Atari-Go Players for Parameter Exploring Policy Gradients

Grüttner M, Sehnke F, Schaul T, Schmidhuber J

International Conference on Artificial Neural Networks (ICANN) (2010)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

A Natural Evolution Strategy for Multi-Objective Optimization

Glasmachers T, Schaul T, Schmidhuber J

Parallel Problem Solving from Nature (PPSN) (2010)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Assessment of neural networks training strategies for histomorphometric analysis of synchrotron radiation medical images

de Moura Meneses A, Pinheiro C, Rancoita P, Schaul T, Gambardella L, Schirru R et al.

Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment (2010)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Exponential Natural Evolution Strategies

Glasmachers T, Schaul T, Sun Y, Wierstra D, Schmidhuber J

Genetic and Evolutionary Computation Conference (GECCO) (2010)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Exploring Parameter Space in Reinforcement Learning

Rückstieß T, Sehnke F, Schaul T, Wierstra D, Sun Y, Schmidhuber J

Paladyn Journal of Behvioral Robotics (2010)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Metalearning

Schaul T, Schmidhuber J

Scholarpedia (2010)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

PyBrain

Schaul T, Bayer J, Wierstra D, Sun Y, Felder M, Sehnke F et al.

Journal of Machine Learning Research (2010)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Frontier Search

Sun Y, Glasmachers T, Schaul T, Schmidhuber J

Conference on Artificial General Intelligence (AGI) (2010)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Towards Practical Universal Search

Schaul T, Schmidhuber J

Conference on Artificial General Intelligence (AGI) (2010)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Curiosity-Driven Optimization

Schaul T, Sun Y, Wierstra D, Gomez F, Schmidhuber J

IEEE Congress on Evolutionary Computation (CEC) (2011)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Stochastic Search using the Natural Gradient

Sun Y, Wierstra D, Schaul T, Schmidhuber J

International Conference on Machine Learning (ICML) (2009)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

A Scalable Neural Network Architecture for Board Games

Schaul T, Schmidhuber J

IEEE Symposium on Computational Intelligence in Games (CIG) (2008)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Countering Poisonous Inputs with Memetic Neuroevolution

Togelius J, Schaul T, Schmidhuber J, Gomez F

Parallel Problem Solving from Nature (PPSN) (2008)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Evolving a compact , concept-based Sokoban solver

Schaul T

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Natural Evolution Strategies

Wierstra D, Schaul T, Peters J, Schmidhuber J

Proceedings of the IEEE Congress on Evolutionary Computation (CEC) (2008)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Efficient Natural Evolution Strategies

Sun Y, Wierstra D, Schaul T, Schmidhuber J

Genetic and Evolutionary Computation Conference (GECCO) (2009)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Fitness Expectation Maximization

Wierstra D, Schaul T, Peters J, Schmidhuber J

Parallel Problem Solving from Nature (PPSN) (2008)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Episodic Reinforcement Learning by Logistic Reward-Weighted Regression

Wierstra D, Schaul T, Peters J, Schmidhuber J

International Conference on Artificial Neural Networks (ICANN) (2008)

  • -- Readers Readers not available. This publication is not currently accessible in the Mendeley catalog.

Professional experience

Research scientist

Google - DeepMind

September 2013 - Present

Research Scientist

NYU Courant Institute of Mathematical Sciences

September 2011 - August 2013 (1 year 11 months)

PhD student

IDSIA

February 2007 - August 2011 (4 years 6 months)

Education history

Technische Universität München

Dr. rer. nat.

February 2007 - August 2011 (4 years 6 months)

EPFL - Swiss Federal Institute of Technology

MSc

October 2000 - May 2005 (4 years 7 months)