We address two fundamental problems in computational and theoretical neuroscience that are not accessible via standard methods based on averaging:
(1) How do single neurons process spike sequences? For regular spike sequences we find a non-monotonic input-output relation such that less input may yield more output. We theoretically derive conditions for this intriguing phenomenon to occur and confirm our results experimentally.
(2) Higher animals as well as bio-mechanically complex robots face the problemof how to control many sensory modalities with many different yet appropriate behaviors in an autonomous way. Here we present a new type of chaos control that we developed to be both adaptive and applicable in neuron-like systems. It is capable of stabilizing a range of different periods at fixed parameters. We implement this adaptive chaos control to replace the commonly used many central pattern generators (CPGs) with intrinsically periodic dynamics by *one* CPG that is intrinsically chaotic but is stabilized to a variety of periodic orbits, depending on the sensory input that controls the CPG. We demonstrate by example control of a hexapod robot that this single-CPG approach makes the robot much more versatile than previously achieved and flexible in its choice of behaviors. Moreover, it can easily learn to associate appropriate behaviors to given environmental conditions.”
Refs.: Nature Physics 6:224 (2010); Nature Physics 6, 161 (2010); Bick, Kolodziejski and Timme, in prep.; Kielblock et al., under review (2011).