D
David Saxton
Researcher at Google
Publications - 30
Citations - 2071
David Saxton is an academic researcher from Google. The author has contributed to research in topics: Hypergraph & Ramsey's theorem. The author has an hindex of 12, co-authored 30 publications receiving 1730 citations. Previous affiliations of David Saxton include University of Cambridge & Instituto Nacional de Matemática Pura e Aplicada.
Papers
More filters
Proceedings Article
Unifying count-based exploration and intrinsic motivation
TL;DR: In this paper, the authors use density models to measure uncertainty and derive a pseudo-count from an arbitrary density model, which can be used to improve exploration in non-tabular reinforcement learning.
Journal ArticleDOI
Hypergraph containers
David Saxton,Andrew Thomason +1 more
TL;DR: The notion of containment for independent sets in hypergraphs was introduced in this article, where it was shown that for a given hypergraph G, there exists a relatively small collection of vertex subsets, such that every independent set of G is contained within a member of the subset, and no member is large; the collection reveals an underlying structure to the independent sets.
Posted Content
Unifying Count-Based Exploration and Intrinsic Motivation
TL;DR: This work uses density models to measure uncertainty, and proposes a novel algorithm for deriving a pseudo-count from an arbitrary density model, which enables this technique to generalize count-based exploration algorithms to the non-tabular case.
Posted Content
Unsupervised Predictive Memory in a Goal-Directed Agent
Greg Wayne,Chia-Chun Hung,David Amos,Mehdi Mirza,Arun Ahuja,Agnieszka Grabska-Barwinska,Jack W. Rae,Piotr Mirowski,Joel Z. Leibo,Adam Santoro,Mevlana Gemici,Malcolm Reynolds,Tim Harley,Josh Abramson,Shakir Mohamed,Danilo Jimenez Rezende,David Saxton,Adam Cain,Chloe Hillier,David Silver,Koray Kavukcuoglu,Matthew Botvinick,Demis Hassabis,Timothy P. Lillicrap +23 more
TL;DR: A model, the Memory, RL, and Inference Network (MERLIN), in which memory formation is guided by a process of predictive modeling, demonstrates a single learning agent architecture that can solve canonical behavioural tasks in psychology and neurobiology without strong simplifying assumptions about the dimensionality of sensory input or the duration of experiences.
Proceedings Article
Conditional Neural Processes
Marta Garnelo,Dan Rosenbaum,Chris J. Maddison,Tiago Ramalho,David Saxton,Murray Shanahan,Yee Whye Teh,Danilo Jimenez Rezende,S. M. Ali Eslami +8 more
TL;DR: Conditional Neural Processes (CNPs) as mentioned in this paper combine the benefits of both deep neural networks and Gaussian Processes by using prior knowledge to quickly infer the shape of a new function at test time.