Abstract
The reshaping and decorrelation of similar activity patterns by neuronal networks can enhance their discriminability, storage, and retrieval. How can such networks learn to decorrelate new complex patterns, as they arise in the olfactory system? Using a computational network model for the dominant neural populations of the olfactory bulb we show that fundamental aspects of the adult neurogenesis observed in the olfactory bulb – the persistent addition of new inhibitory granule cells to the network, their activity-dependent survival, and the reciprocal character of their synapses with the principal mitral cells – are sufficient to restructure the network and to alter its encoding of odor stimuli adaptively so as to reduce the correlations between the bulbar representations of similar stimuli. The decorrelation is quite robust with respect to various types of perturbations of the reciprocity. The model parsimoniously captures the experimentally observed role of neurogenesis in perceptual learning and the enhanced response of young granule cells to novel stimuli. Moreover, it makes specific predictions for the type of odor enrichment that should be effective in enhancing the ability of animals to discriminate similar odor mixtures.
Highlights
Contrast enhancement and decorrelation are common steps in information processing
The olfactory bulb is one of only two brain regions in which new neurons are added persistently in substantial numbers even in adult animals. This leads to an ongoing turnover of interneurons, in particular of the inhibitory granule cells, which constitute the largest cell population of the olfactory bulb
We present a basic computational model that is built on fundamental aspects of the granule cells and their connections with the excitatory mitral cells, which convey the olfactory information to higher brain areas
Summary
Contrast enhancement and decorrelation are common steps in information processing. They can reshape neuronal activity patterns so as to enhance down-stream processing like pattern discrimination, storage, and retrieval. It does so despite the fact that even simple odors evoke complex activation patterns due to the fractured representation of the high-dimensional odor space on the twodimensional glomerular surface [4]. Unlike spatial contrast enhancement in the retina [5], this decorrelation can not arise from local lateral inhibition that is confined to neighboring glomeruli [3,6]. What types of network connectivities can underlie the enhancement of small, but significant differences in the representation of similar odors?
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
