S. Klampfl, R. Legenstein, and W. Maass
The extraction of statistically independent components from high-dimensional
multi-sensory input streams is assumed to b e an essential component of
sensory processing in the brain. Such independent component analysis (or
blind source separat ion) could provide a less redundant representation of
inform ation about the external world. Another powerful processing strategy
is to extract preferentially those components from high-dimensional input
streams that are related to other inf ormation sources, such as internal
predictions or propriocep tive feedback. This strategy allows the
optimization of inte rnal representation according to the information
bottleneck method. However, concrete learning rules that implement thes e
general unsupervised learning principles for spiking neuro ns are still
missing. We show how both information bottlenec k optimization and the
extraction of independent components can in principle be implemented with
stochastically spiking neurons with refractoriness. The new learning rule
that achi eves this is derived from abstract information optimization
principles.