Optimization of a dolphin hearing model to relative sensitivity and frequency discrimination through simple aggregate selection

Abstract

Dolphins demonstrate a broad range of hearing (>120 kHz), a remarkable frequency-dependent auditory sensitivity, and a fine degree of frequency discrimination across the range of hearing. Evolutionary programming has been successfully used to model the relative auditory sensitivity of the dolphin, but similar models of frequency discrimination have not been attempted. Evolutionary programming is used to create a bandpass ear model of the dolphin by optimizing output to the observed frequency discrimination ability of the dolphin. By implementing simple aggregate selection, a similar approach is used to create a model optimized to both the frequency discrimination ability and the relative auditory sensitivity of the dolphin. Success at modeling the frequency discrimination abilities of the dolphin holds promise for future work in developing biologically realistic models of dolphin hearing. However, the inability of the evolutionary program to optimize both frequency discrimination and relative auditory sensitivity suggests that simple aggregate selection may not provide a sufficiently rigorous performance metric for dual optimization, or that the problem is intractable within the constraints of the current model design.