Society for Mathematical Biology nautilus logo

International Conference on Mathematical Biology and

Annual Meeting of The Society for Mathematical Biology,

July 27-30, 2009

University of British Columbia, Vancouver

.

Program

Poster PS26A
Edward K. Agarwala
Case Western Reserve University
Title Food for Thought: When Infomax Fails to Optimize Utility
Abstract Information maximization criteria have been used to account for the physiology of sensory systems as diverse as receptive fields in the primary visual and auditory cortices, the organization of saccades by the oculomotor system, and olfaction. We investigated a simple model of an organism searching for food by taking successive samples from an environment in which food particles diffuse stochastically from a slowly randomly moving source. In the limit of large food concentrations we reduced our high dimensional model system to a Markov chain on a small number (<= 5) of equivalence classes on the state space. In this system were are able to make rigorous quantitative comparisons of the relative benefit to the organism of pursuing a search strategy based on (i) maximizing the searcher's information about the location of the food source, (ii) maximizing the expected concentration of food at the searcher's next location, and (iii) hybrid strategies combining aspects of (i) and (ii). The relative success of the different strategies in terms of long-term expected food benefit depended on parameters such as the variability of the source movement. To our surprise we found that each strategy was superior to the others for a certain set of parameters. Therefore any claim that the principles of information maximization or immediate utility optimization provide an explanation of creature behavior should be met with suspicion.
CoauthorsPeter J. Thomas
LocationWoodward Lobby (Monday-Tuesday)