|
| ||
Natural Scene Statistics and the Evolution of Perceptual Systems | ||
W.S. Geisler and R.L. Diehl | ||
University of Texas at Austin | ||
Recent advances in measuring natural scene statistics raise the possibility of developing more complete and testable theories of evolution by combining precise statistical descriptions of the environment with precise statistical descriptions of genetics. We propose a formal framework for analyzing how the statistics of natural stimuli and the process of natural selection interact to determine the design of perceptual systems. The framework consists of two parts. One is a Bayesian ideal observer with a utility function appropriate for natural selection. The other is a Bayesian formulation of natural selection. In the Bayesian formulation of natural selection, each allele vector (polymorphism) in each species under consideration is represented by a fundamental equation, which describes how the number of organisms carrying that allele vector at time t+1 is related to (1) the number of organisms carrying that allele vector at time t, (2) the prior probability of a state of the environment at time t, (3) the likelihood of a stimulus given the state of the environment, (4) the likelihood of a response given the stimulus, and (5) the birth and death rates given the response and the state of the environment. The process of natural selection is represented by iterating these fundamental equations in parallel over time, while updating the allele vectors using appropriate probability distributions for mutation and sexual recombination. We show that simulations of Bayesian natural selection can yield new insights, for example, into the co-evolution of camouflage and color vision. We believe that Bayesian natural selection offers an appropriate theoretical framework for investigating perceptual systems, as well as many other biological systems. | ||