Information distortion as a model of sensory processing
Alex Dimitrov, Department of Cell Biology and Neuroscience, Montana State University
The information distortion analysis provides a view of sensory
processing that is essentially one of signal discrimination: There are
a few signal targets (classes) that are recognized. They are
represented by a few response types (classes). However, evidence from
behavioral and anatomical studies suggest that there are additional
quantities, complementary to the discrimination problem (direction,
velocity, distance to objects). Such quantities appear as symmetries
to the discrimination problem. Typically, approaches to sensory
processing disregard these symmetries (as in invariant object
discrimination, ``what'' visual pathway), or relegate their handling
to a separate subsystem (e.g., ``where'' visual pathway). I will
discuss two modifications to the information distortion method that
takes the symmetries into account: an explicit approach, which
includes the symmetries in the quantization procedure, and an implicit
approach, which represents the action of the symmetry groups as a set
of unrelated clusters.