Masterthesis: Opponent Models in Bilateral Negotiation

by Mark Hendrikx.

Literature Survey

The newst version can be found on the ​SVN repository

Main Research Questions

• How to measure the quality of opponent models?
• Are agents using opponent models "better" than other agents?
• How do opponent models compare regarding accuracy, convergence, and computational cost?
• How to create an opponent model which is better than the state-of-the-art?

Meeting

​Meeting agendas and minutes

Planning

The datas in the overview below refer to deadlines.

• Validated all decoupled agents (13-01-12)

• Study for exam (15-01-12 -> 03-02-12)

• Quantified the benefits of using a perfect model (07-02-12)

• Finished editing OM paper (15-02-12)
  • Abstract / Introduction (08-01-12)
  • Preliminaries (09-01-12)
  • OM in automated negotiation (10-01-12)
  • OM Framework (11-01-12)
  • Learning of OM (07-02-12)
  • Use of OM (09-02-12)
  • Evaluating OM (10-02-12)
  • OM in ANAC (11-02-12)
  • Quality of OM in ANAC (13-02-12)
  • Conclusion (15-02-12)

• First version of short paper ACAN (18-02-12)

• Compare accuracy, convergence, and computational cost of existing models (30-03-12)
  • Implement measurement methods
  • Create testsetup
  • Analyze data

• Finished ANAC2012 agent (18-03-12)

• Create a model which is beter than the state of the art
  • Improve frequency model by dynamic learning rate
  • Improve Bayesian model by improving decision function assumption

• Draft of masterthesis (10/06/12)

• Finish masterthesis (10/07/12)