Sparse spike coding in an asynchronous feed-forward multi-layer neural network using Matching Pursuit
This paper introduced for the first time the use of Matching Pursuit in a neuromimetic architecture of low-level visual areas. It aims at understanding the efficiency and generality of correlation-based inhibition propagation, by showing:
- the link between sparse coding, spike coding and matching pursuit,
the use of a quantization in Matching Pursuit (see Fig. 1),
that a Sparse Hebbian Learning scheme could be applied, leading to the learning of edge-like filter (this is the first report of using an adaptive matching pursuit scheme to my knowledge) -- this was developped in (Perrinet, 03, IEEE) and (Fischer, 07)),
- the possible use as a model of low-level saliency.
Figure 2: (A) Look-Up-Table : mean and variance of the absolute contrast in function of the relative rank for a database of 100 natural images (B) Comparison of MSE in function of the file size (in ) for the different methods : (a) MP; (b) MP with Look-Up-Table; (c) JPEG at different qualities.
Get a preprint
- Laurent Perrinet, Manuel Samuelides, Simon Thorpe. Sparse spike coding in an asynchronous feed-forward multi-layer neural network using Matching Pursuit., URL . Neurocomputing, 57C:125--34, 2002, abstract
All material (c) L. Perrinet. Please check the copyright notice.