Active Learning-Based Identification of Neuronal Assemblies in Parallel Spike Trains
| Type of publication: | Inproceedings |
| Citation: | |
| Booktitle: | Proceedings. 24. Workshop Computational Intelligence, Dortmund, 27.-28. November 2014 |
| Year: | 2014 |
| Month: | November |
| Pages: | 155-172 |
| Publisher: | KIT Scientific Publishing |
| Abstract: | For understanding how information is processed within the brain several different models have been proposed. They are either based on a common increase in neuronal firing activity or synchronous firing of several individual neurons. We present a novel method for detecting the so-called assemblies of the latter model. Using parallel spike trains – recordings of neuronal activity – one use this information to answer quite well if an individual neuron belongs to (at least) one assembly or not. But detecting the underlying assembly structure remains a difficult task since normally neither the number of assemblies is known nor their respective size. Using surrogate-based statistics as an oracle we use active learning to identify the underlying assembly structure. This approach not only uses the statistical information we calculate from the surrogates but the structural information we can obtain by defining a metric on the spike trains themselves. We show that for even a small number of coincidences relatively small assemblies can be detected without querying the oracle for every spike train. |
| Keywords: | |
| Authors | |
| Editors | |
| Added by: | [] |
| Total mark: | 0 |
|
Attachments
|
|
|
Notes
|
|
|
|
|
|
Topics
|
|
