Timezone: »
Inspired by recent work in attention models for image captioning and question answering, we present a soft attention model for the reinforcement learning domain. This model bottlenecks the view of an agent by a soft, top-down attention mechanism, forcing the agent to focus on task-relevant information by sequentially querying its view of the environment. The output of the attention mechanism allows direct observation of the information used by the agent to select its actions, enabling easier interpretation of this model than of traditional models. We analyze the different strategies the agents learn and show that a handful of strategies arise repeatedly across different games. We also show that the model learns to query separately about space and content (where'' vs.
what'').
We demonstrate that an agent using this mechanism can achieve performance competitive with state-of-the-art models on ATARI tasks while still being interpretable.
Author Information
Alexander Mott (DeepMind)
Daniel Zoran (DeepMind)
Mike Chrzanowski (Google Brain)
Daan Wierstra (DeepMind Technologies)
Danilo Jimenez Rezende (Google DeepMind)
More from the Same Authors
-
2021 : Systematic Evaluation of Causal Discovery in Visual Model Based Reinforcement Learning »
Nan Rosemary Ke · Aniket Didolkar · Sarthak Mittal · Anirudh Goyal · Guillaume Lajoie · Stefan Bauer · Danilo Jimenez Rezende · Yoshua Bengio · Chris Pal · Michael Mozer -
2021 : Implicit Riemannian Concave Potential Maps »
Danilo Jimenez Rezende · Sébastien Racanière -
2021 : Implicit Riemannian Concave Potential Maps »
Danilo Jimenez Rezende · Sébastien Racanière -
2021 : Implicit Riemannian Concave Potential Maps »
Danilo Jimenez Rezende · Sébastien Racanière -
2021 Poster: SIMONe: View-Invariant, Temporally-Abstracted Object Representations via Unsupervised Video Decomposition »
Rishabh Kabra · Daniel Zoran · Goker Erdogan · Loic Matthey · Antonia Creswell · Matt Botvinick · Alexander Lerchner · Chris Burgess -
2019 : Equivariant Hamiltonian Flows »
Danilo Jimenez Rezende -
2019 : Morning Coffee Break & Poster Session »
Eric Metodiev · Keming Zhang · Markus Stoye · Randy Churchill · Soumalya Sarkar · Miles Cranmer · Johann Brehmer · Danilo Jimenez Rezende · Peter Harrington · AkshatKumar Nigam · Nils Thuerey · Lukasz Maziarka · Alvaro Sanchez Gonzalez · Atakan Okan · James Ritchie · N. Benjamin Erichson · Harvey Cheng · Peihong Jiang · Seong Ho Pahng · Samson Koelle · Sami Khairy · Adrian Pol · Rushil Anirudh · Jannis Born · Benjamin Sanchez-Lengeling · Brian Timar · Rhys Goodall · Tamás Kriváchy · Lu Lu · Thomas Adler · Nathaniel Trask · Noëlie Cherrier · Tomohiko Konno · Muhammad Kasim · Tobias Golling · Zaccary Alperstein · Andrei Ustyuzhanin · James Stokes · Anna Golubeva · Ian Char · Ksenia Korovina · Youngwoo Cho · Chanchal Chatterjee · Tom Westerhout · Gorka Muñoz-Gil · Juan Zamudio-Fernandez · Jennifer Wei · Brian Lee · Johannes Kofler · Bruce Power · Nikita Kazeev · Andrey Ustyuzhanin · Artem Maevskiy · Pascal Friederich · Arash Tavakoli · Willie Neiswanger · Bohdan Kulchytskyy · sindhu hari · Paul Leu · Paul Atzberger -
2019 : Panel »
Sanja Fidler · Josh Tenenbaum · Tatiana López-Guevara · Danilo Jimenez Rezende · Niloy Mitra -
2019 : Danilo Rezende »
Danilo Jimenez Rezende -
2019 Poster: Shaping Belief States with Generative Environment Models for RL »
Karol Gregor · Danilo Jimenez Rezende · Frederic Besse · Yan Wu · Hamza Merzic · Aaron van den Oord -
2018 Poster: A Probabilistic U-Net for Segmentation of Ambiguous Images »
Simon Kohl · Bernardino Romera-Paredes · Clemens Meyer · Jeffrey De Fauw · Joseph R. Ledsam · Klaus Maier-Hein · S. M. Ali Eslami · Danilo Jimenez Rezende · Olaf Ronneberger -
2018 Spotlight: A Probabilistic U-Net for Segmentation of Ambiguous Images »
Simon Kohl · Bernardino Romera-Paredes · Clemens Meyer · Jeffrey De Fauw · Joseph R. Ledsam · Klaus Maier-Hein · S. M. Ali Eslami · Danilo Jimenez Rezende · Olaf Ronneberger -
2018 Poster: Relational recurrent neural networks »
Adam Santoro · Ryan Faulkner · David Raposo · Jack Rae · Mike Chrzanowski · Theophane Weber · Daan Wierstra · Oriol Vinyals · Razvan Pascanu · Timothy Lillicrap -
2017 Poster: Imagination-Augmented Agents for Deep Reinforcement Learning »
Sébastien Racanière · Theophane Weber · David Reichert · Lars Buesing · Arthur Guez · Danilo Jimenez Rezende · Adrià Puigdomènech Badia · Oriol Vinyals · Nicolas Heess · Yujia Li · Razvan Pascanu · Peter Battaglia · Demis Hassabis · David Silver · Daan Wierstra -
2017 Oral: Imagination-Augmented Agents for Deep Reinforcement Learning »
Sébastien Racanière · Theophane Weber · David Reichert · Lars Buesing · Arthur Guez · Danilo Jimenez Rezende · Adrià Puigdomènech Badia · Oriol Vinyals · Nicolas Heess · Yujia Li · Razvan Pascanu · Peter Battaglia · Demis Hassabis · David Silver · Daan Wierstra -
2017 Poster: Variational Memory Addressing in Generative Models »
Jörg Bornschein · Andriy Mnih · Daniel Zoran · Danilo Jimenez Rezende -
2017 Poster: Visual Interaction Networks: Learning a Physics Simulator from Video »
Nicholas Watters · Daniel Zoran · Theophane Weber · Peter Battaglia · Razvan Pascanu · Andrea Tacchetti -
2016 Poster: Unsupervised Learning of 3D Structure from Images »
Danilo Jimenez Rezende · S. M. Ali Eslami · Shakir Mohamed · Peter Battaglia · Max Jaderberg · Nicolas Heess -
2016 Poster: Towards Conceptual Compression »
Karol Gregor · Frederic Besse · Danilo Jimenez Rezende · Ivo Danihelka · Daan Wierstra -
2016 Poster: Interaction Networks for Learning about Objects, Relations and Physics »
Peter Battaglia · Razvan Pascanu · Matthew Lai · Danilo Jimenez Rezende · koray kavukcuoglu -
2016 Poster: Matching Networks for One Shot Learning »
Oriol Vinyals · Charles Blundell · Timothy Lillicrap · koray kavukcuoglu · Daan Wierstra -
2015 Poster: Variational Information Maximisation for Intrinsically Motivated Reinforcement Learning »
Shakir Mohamed · Danilo Jimenez Rezende -
2014 Poster: Shape and Illumination from Shading using the Generic Viewpoint Assumption »
Daniel Zoran · Dilip Krishnan · José Bento · Bill Freeman -
2014 Poster: Semi-supervised Learning with Deep Generative Models »
Diederik Kingma · Shakir Mohamed · Danilo Jimenez Rezende · Max Welling -
2014 Spotlight: Semi-supervised Learning with Deep Generative Models »
Diederik Kingma · Shakir Mohamed · Danilo Jimenez Rezende · Max Welling -
2013 Poster: Learning the Local Statistics of Optical Flow »
Dan Rosenbaum · Daniel Zoran · Yair Weiss