Regensburg 2002 – scientific programme
Parts | Days | Selection | Search | Downloads | Help
SYPH: Physik im Hirn - Physical Approaches to Brain Function
SYPH 2: Physik im Hirn - Physical Approaches to Brain Function
SYPH 2.3: Poster
Thursday, March 14, 2002, 15:30–18:00, Poster D
Optimal neuronal population coding and Fisher information — •Matthias Bethge, David Rotermund, and Klaus Pawelzik — University of Bremen, Institute of Theoretical Physics,
Kufsteiner Str., D-28334 Bremen
Efficient coding has been proposed as a first principle explaining neuronal response properties in the central nervous system (Barlow). Optimal codes, however, depend on the natural limitation of the particular physical system. Here we investigate how optimal neural encoding strategies are influenced by the finite number of neurons N (place constraint), the limited decoding time window length T (time constraint), the maximum neuronal firing rate fmax (power constraint) and the maximal average rate ⟨ f ⟩max (energy constraint). While Fisher information provides a general lower bound for the mean squared error of unbiased signal reconstruction, its use for determining optimal encoding strategies is limited. Analyzing simple examples, we illustrate some typical pitfalls and thereby show that Fisher information mainly provides a valid measure for the precision of a code in the low-noise limit given by large observation time windows. In particular, we show that Fisher information is not suitable to derive characteristic signatures of optimal population codes. Alternatively, we analyze the minimum mean squared error, whereby it turns out that labeled line coding constitutes the major contribution to population codes for small T and large N, while intensity coding, quantified by Fisher information, can be neglected. Furthermore, it turns out that the accuracy of unimodal coding schemes depends only weakly on the tuning width as long as energy consumption can be neglected.