Application of Negative Correlation in Neural Information Processing and Image Reconstruction

Negative correlation has clear statistical benefits for noise reduction and data representation. This talk is to introduce two recent papers. In the first paper, suprathreshold stochastic resonance is examined in the context of integrate-and-fire neurons, with an emphasis on the optimality of negative correlation in the neuronal firing. In the second paper, two new algorithms, correlated component analysis and correlated basis analysis are designed to exploit the benefits of negative correlation, which extend the approach of independent component analysis.

Venue and Time/Date

CS.101 at 1pm on Tue, 5th of July 2011