The Coconut Model with Adaptive Strategies
We continue our development of an agent-based baseline model for the Coconut Model by Peter Diamond and report on the achievements.
While dealing with fixed but heterogeneous strategies in the Talk presented last year, we now concentrate on learning dynamics.
Namely, we introduce temporal difference learning as a way to procedurally solve the optimization problem as posed in the original paper.
We show that the model with this kind of adaptive agents converges to a considerable degree to the original theoretical results for an infinite and homogeneously adapting population in terms of learning dynamics alone as well as regarding the overall model behavior.
Conclusions regarding non-equilibrium trajectories and equilibrium selection can be drawn from that.