Colour versus Shape Goal Misgeneralization in Reinforcement Learning: A Case Study

1 code implementation • 5 Dec 2023 • Karolis Ramanauskas, Özgür Şimşek

We explore colour versus shape goal misgeneralization originally demonstrated by Di Langosco et al. (2022) in the Procgen Maze environment, where, given an ambiguous choice, the agents seem to prefer generalization based on colour rather than shape.

reinforcement-learning

Creating Multi-Level Skill Hierarchies in Reinforcement Learning

1 code implementation • NeurIPS 2023 • Joshua B. Evans, Özgür Şimşek

What is a useful skill hierarchy for an autonomous agent?

graph partitioning reinforcement-learning

Explaining Reinforcement Learning with Shapley Values

1 code implementation • 9 Jun 2023 • Daniel Beechey, Thomas M. S. Smith, Özgür Şimşek

For reinforcement learning systems to be widely adopted, their users must understand and trust them.

reinforcement-learning

Resource-Constrained Station-Keeping for Helium Balloons using Reinforcement Learning

no code implementations • 2 Mar 2023 • Jack Saunders, Loïc Prenevost, Özgür Şimşek, Alan Hunter, Wenbin Li

Very recently, reinforcement learning has been proposed as a control scheme to maintain the balloon in the region of a fixed location, facilitated through diverse opposing wind-fields at different altitudes.

Continuous Control Navigate +2

Iterative Policy-Space Expansion in Reinforcement Learning

no code implementations • 5 Dec 2019 • Jan Malte Lichtenberg, Özgür Şimşek

Humans and animals solve a difficult problem much more easily when they are presented with a sequence of problems that starts simple and slowly increases in difficulty.

reinforcement-learning Reinforcement Learning (RL)

The Game of Tetris in Machine Learning

1 code implementation • 5 May 2019 • Simón Algorta, Özgür Şimşek

The game of Tetris is an important benchmark for research in artificial intelligence and machine learning.

BIG-bench Machine Learning reinforcement-learning +1

Learning From Small Samples: An Analysis of Simple Decision Heuristics

no code implementations • NeurIPS 2015 • Özgür Şimşek, Marcus Buckmann

Simple decision heuristics are models of human and animal behavior that use few pieces of information---perhaps only a single piece of information---and integrate the pieces in simple ways, for example, by considering them sequentially, one at a time, or by giving them equal weight.

Decision Making

Linear decision rule as aspiration for simple decision heuristics

no code implementations • NeurIPS 2013 • Özgür Şimşek

Many attempts to understand the success of simple decision heuristics have examined heuristics as an approximation to a linear decision rule.

Skill Characterization Based on Betweenness

no code implementations • NeurIPS 2008 • Özgür Şimşek, Andrew G. Barto

We present a characterization of a useful class of skills based on a graphical representation of an agent's interaction with its environment.