1: \begin{abstract}

2: 	\noindent Spike trains data find a growing list of applications in computational 

3: 	neuroscience,

4: 	imaging, streaming data and finance. Machine learning strategies for spike

5: 	trains are based on various neural network and probabilistic models. The probabilistic

6: 	approach is relying on parametric or nonparametric specifications of the underlying spike

7: 	generation model. In this paper we

8: 	consider the two-class statistical classification problem for a class of spike train data

9: 	characterized by nonparametrically specified intensity functions. We derive the optimal

10: 	Bayes rule and next form the plug-in nonparametric kernel classifier. Asymptotical

11: 	properties of the rules are established including the limit with respect to the

12: 	increasing recording time interval and the size of a training set.

13: 	In particular the convergence of the kernel classifier to the Bayes rule is proved.

14: 	The obtained results are supported by a finite sample simulation studies.

15: \end{abstract}

16: