Frontiers in Computational Neuroscience will be publishing an interesting paper written by Honi Sanders and John Lisman (with co-authors Brian E. Kolterman, Roman Shusterman, Dmitry Rinberg, Alexei Koulakov) titled, “A network that performs brute-force conversion of a temporal sequence to a spatial pattern: relevance to odor recognition“. Honi Sanders has written a preview of this paper.
by Honi Sanders
There are many occasions in which the brain needs to process information that is provided in a sequence. These sequences may be externally generated or internally generated. For example, in the case of understanding speech, where words that come later may affect the meaning of words that come earlier, the brain must somehow store the sentence it is receiving long enough to process the sentence as a whole. On the other hand, sequences of information also are passed from one brain area to another. In these cases too the brain must store the sequence it is receiving long enough to process the message as a whole.
One such sequence is generated by the olfactory bulb, which is the second stage of processing of the sense of smell. While individual cells in the olfactory bulb will fire bursts in response to many odors, the order in which they fire is specific to an individual odor. How such a sequence can be recognized as a specific odor remains unclear. In Sanders et al, we present experimental evidence that the sequence is discrete and therefore contains a relatively small number of sequential elements; each element is represented in a given cycle of the gamma frequency oscillations that occur during a sniff. This raises the possibility of a “brute force” solution for converting the sequence into a spatial pattern of the sort that could be recognized by standard “attractor” neural networks. We present computer simulations of model networks that have modules; each model can produce a persistent snapshot of what occurs during a given gamma cycle. In this way, the unique properties of the sequence can be determined at the end of sniff by the spatial pattern of cell firing in all modules.
The authors thank Brandeis University High Performance Computing Cluster for cluster time. This work was supported by the NSF Collaborative Research in Computational Neuroscience, NSF IGERT, and the Howard Hughes Medical Institute.