Neural networks have become a fertile meeting ground for biologists, physicists, and computer scientists. Studies of surprisingly skilled animal behavior have challenged physicists to explain sensory capabilities that seem to exceed the physical limitations of sense organs and neural interactions. For example, a barn owl at night deduces the direction to an unsuspecting mouse by perceiving the interaural arrival-time difference of its rustling with microsecond accuracy (see Physics Today, June 2001, page 20). But how can that be when the characteristic time of an individual neuronal process is 100 times slower?
A new paper in Physical Review Letters by biophysicist Leo van Hemmen and colleagues at the Technical University of Munich proposes a neural-network model that addresses a similar problem raised by the spatial acuity of infrared imaging by certain kinds of snakes. 1 Ten years ago, van Hemmen’s group, which specializes in the theory of biosensory systems...
