Univ. of Essex (Local organizer), Univ. of Giessen, Univ. of Cyprus
Motivation and objectives
Agent based computational economic (ACE) models are increasingly recognized as a valuable tool in analysing complex systems. The tutorial introduces the participants to ACE. After a general presentation, two main applications of ACE will be developed: (i) computational mechanism design of market micro-structure and auction design, and (ii) hypothesis testing of phenomena from artificial market environments, e.g., artificial stock market (ASM) models. Heuristic and AI methods will be used for robustness analysis of proposed auction/market protocols and to simulate learning and decision making behaviour of market. Heuristics will also be used to estimate model parameters and fine tune them to fit the properties of real markets.
Two days with six lectures of 90 minutes and an afternoon "hands on" session in the computer laboratory.
- Markets as complex artificial systems: Principles of ABM; self organization; dynamics from large numbers of interacting agents; examples.
- Market and auction design: Trading platforms; smart markets for congestion; computational test bedding/“wind tunnel tests” with Erev-Roth type reinforcement learning on how to play games strategically.
- Artificial stock markets: Selected models for simulating stock markets and their properties
- Reinforced learning principle: Applications in financial modelling; implementation of stock trading agents with adaptive behaviour.
- ASM hypothesis testing: Stylized facts of popular artificial stock market models; parameter estimation techniques; heuristic parameter estimation.
- Advanced issues in ASM modelling.
- "Hands-on" session: implementation of an artificial stock market model; application of heuristic search techniques to estimating model parameters for empirical data.