Adaptive and Learning Agents

Volume 5924 of the series Lecture Notes in Computer Science pp 107-122

Learning to Locate Trading Partners in Agent Networks

  • John PorterAffiliated withDepartment of Computer Science, University of Tulsa
  • , Kuheli ChakrabortyAffiliated withDepartment of Computer Science, University of Tulsa
  • , Sandip SenAffiliated withDepartment of Computer Science, University of Tulsa

* Final gross prices may vary according to local VAT.

Get Access


This paper is motivated by some recent, intriguing research results involving agent-organized networks (AONs). In AONs agents have a limited number of collaboration partners at any time, represented by edges in a network of agent nodes, and can rewire edges, i.e., change partners, to improve performance. The common underlying research issue in these domains is the search for desirable interaction or collaboration partners in a relatively large population. Agents have to learn to estimate the utility of current trading partners and adapt connections to improve profitability. A previous study found that random selection of partners in each time period produced better performance but incurred larger search costs compared to gradual rewiring of edges in the network in a production and exchange economy. We propose an exponentially decaying exploration scheme that produces similar utilities to random rewiring but with much less rewiring costs. We evaluate the effects of the number of trading partners on the utilities obtained by the agents. We hypothesize on the cause for the observed performance differences and verify that by showing that the observed performance differences with more realistic model of the economy that incorporate minimum trade volumes and storage capacities.


Agent oriented networks partner selection learning