Synergy of stochastic and nature inspired optimization in recommender systems

Abstract

In this extended abstract we describe experiences in trying to combine and/or coordinate results of two or more recommendation methods to produce a joint effect greater than the sum of their separate effect - i.e. synergize. We present an overview of results in several directions, lessons learned and some experiments which will guide future research.

Our main tool is a dynamic island platform with different stochastic and evolutionary optimization algorithms with migration managed by a multi-agent planner. We present initial results on MovieLens datasets.

Keywords

• evolutionary island model
• neural networks
• multi-agent planner