Collaborative Expert Portfolio Management

Authors

  • David Stern Microsoft FUSE Labs
  • Horst Samulowitz National ICT Australia and University of Melbourne
  • Ralf Herbrich Microsoft FUSE Labs
  • Thore Graepel Microsoft Research
  • Luca Pulina Universita di Genova
  • Armando Tacchella Universita di Genova

DOI:

https://doi.org/10.1609/aaai.v24i1.7561

Keywords:

Algorithm Portfolios, Recommender System, Machine Learning, Expert Portfolios, QBF, Linear Programming

Abstract

We consider the task of assigning experts from a portfolio of specialists in order to solve a set of tasks. We apply a Bayesian model which combines collaborative filtering with a feature-based description of tasks and experts to yield a general framework for managing a portfolio of experts. The model learns an embedding of tasks and problems into a latent space in which affinity is measured by the inner product. The model can be trained incrementally and can track non-stationary data, tracking potentially changing expert and task characteristics. The approach allows us to use a principled decision theoretic framework for expert selection, allowing the user to choose a utility function that best suits their objectives. The model component for taking into account the performance feedback data is pluggable, allowing flexibility. We apply the model to manage a portfolio of algorithms to solve hard combinatorial problems. This is a well studied area and we demonstrate a large improvement on the state of the art in one domain (constraint solving) and in a second domain (combinatorial auctions) created a portfolio that performed significantly better than any single algorithm.

Downloads

Published

2010-07-03

How to Cite

Stern, D., Samulowitz, H., Herbrich, R., Graepel, T., Pulina, L., & Tacchella, A. (2010). Collaborative Expert Portfolio Management. Proceedings of the AAAI Conference on Artificial Intelligence, 24(1), 179-184. https://doi.org/10.1609/aaai.v24i1.7561

Issue

Section

Constraints, Satisfiability, and Search