From Complexity to Agent Modelling and Back Again - Bruce Edmonds
The difficulty of making correct predictions about its environment (measured by its error rate) for an agent using the best model it can infer from the information available to it given its computational resources.This is a more precise version of the "economic complexity" described in section 1.1. We will define the "model complexity", pertaining to a particular model of an agent as:
The computational difficulty of reaching and testing a model given the constraints of the language and the known data it has to fit.I will consider four situations representing cases of increasing environmental complexity. The environmental complexity will affect how the agent needs to evaluate its models, in particular the kind of trade-offs between the model complexity, specificity and error.
Generated with CERN WebMaker