On ensemble techniques for AIXI approximation

Authors:
Joel Veness;Peter Sunehag;Marcus Hutter
Affiliations:
University of Alberta, Canada;Australian National University, Australia;Australian National University, Australia
Venue:
AGI'12 Proceedings of the 5th international conference on Artificial General Intelligence
Year:
2012

Citing 11
Cited 0

Tracking the Best Expert

Machine Learning - Special issue on context sensitivity and concept drift
Introduction to Reinforcement Learning

Introduction to Reinforcement Learning
Switching Between Two Universal Source Coding Algorithms

DCC '98 Proceedings of the Conference on Data Compression
Convex Optimization

Convex Optimization
Universal Artificial Intelligence: Sequential Decisions Based On Algorithmic Probability

Universal Artificial Intelligence: Sequential Decisions Based On Algorithmic Probability
Efficient algorithms for online convex optimization and their applications

Efficient algorithms for online convex optimization and their applications
A Monte-Carlo AIXI approximation

Journal of Artificial Intelligence Research
Logarithmic regret algorithms for online convex optimization

COLT'06 Proceedings of the 19th annual conference on Learning Theory
Context Tree Switching

DCC '12 Proceedings of the 2012 Data Compression Conference
Mixing Strategies in Data Compression

DCC '12 Proceedings of the 2012 Data Compression Conference
The context-tree weighting method: basic properties

IEEE Transactions on Information Theory

Quantified Score

Hi-index	0.00

Visualization

Abstract

One of the key challenges in AIXI approximation is model class approximation - i.e. how to meaningfully approximate Solomonoff Induction without requiring an infeasible amount of computation? This paper advocates a bottom-up approach to this problem, by describing a number of principled ensemble techniques for approximate AIXI agents. Each technique works by efficiently combining a set of existing environment models into a single, more powerful model. These techniques have the potential to play an important role in future AIXI approximations.