Summary
International Symposium on Nonlinear Theory and Its Applications
2015
Session Number:B3L-F
Session:
Number:B3L-F-5
Verification of Parameter Estimation Techniques from Spike Train Data
Huu Hoang, Isao T. Tokuda,
pp.648-651
Publication Date:2015/12/1
Online ISSN:2188-5079
Summary:
The inverse problem of estimating parameters from spike trains needs a stochastic approach to find most likely solutions. Since the brain exhibits complicated dynamics that is difficult for the model to reproduce, the modeling errors are inevitable. In our recent study, we proposed a Bayesian framework to estimate two model parameters in a segment-wise fashion and then to merge the segmental estimates into a single estimate. The segmental Bayes compensated the modeling errors caused by the mismatch between the brain and the model. The previous study, however, has not yet been properly validated, because it was applied to experimental data, the true parameter values of which are unknown. The aim of this paper is to evaluate the segmental Bayes using simulated spike data, for which the true parameter values are known. The performance evaluation confirmed that the segmental Bayes outperforms other approaches. It also has a strong robustness against non-stationarity of the spike data. We thus conclude that the segmental Bayes provides a useful tool in neuroscience to estimate parameters from spike trains.