What is the goal of sensory coding?
Neural Computation
Independent component analysis, a new concept?
Signal Processing - Special issue on higher order statistics
The nature of statistical learning theory
The nature of statistical learning theory
CNS '97 Proceedings of the sixth annual conference on Computational neuroscience : trends in research, 1998: trends in research, 1998
Blind Source Separation by Sparse Decomposition in a Signal Dictionary
Neural Computation
Do Simple Cells in Primary Visual Cortex Form a Tight Frame?
Neural Computation
Computers in Biology and Medicine
Covert attention with a spiking neural network
ICVS'08 Proceedings of the 6th international conference on Computer vision systems
Role of homeostasis in learning sparse representations
Neural Computation
| Hi-index | 0.00 |
As an alternative to classical representations in machine learningalgorithms, we explore coding strategies using events as is observed forspiking neurons in the central nervous system. Focusing on visualprocessing, we have previously shown that we can define with anover-complete dictionary a sparse spike coding scheme byimplementing lateral interactions that account for redundantinformation. Since this class of algorithms is both compatible withbiological constraints and with neuro-physiological observations, it canprovide a possible algorithm to explain the speed of visual processingdespite the relatively slow time of response of single neurons. Here, Iexplore learning mechanisms to derive in an unsupervised manner anover-complete set of filters which provides a progressively sparserrepresentation of the input. This work is based on a previous model ofsparse coding from Olshausen et al. (1998) and the resultsleads to similar results, suggesting that this strategy provides asimple neural implementation of this algorithm and thus of Blind SourceSeparation. Moreover, this neuro-mimetic algorithm may be easilyextended to realistic architectures of cortical columns in the primaryvisual cortex and we show results for different strategies ofrepresentation, leading to neuro-mimetic adaptive sparse spikecoding schemes.