Interpretation of Inter-spike Interval Statistics through the Markov Switching Poisson Process
Intcr-npikc internal fttatwtic.s arc often uaed to characterize spike sequenccs. Each of lower order statistieal coefficients itself characterize a npikr xtxptrncc well. But it w hard to understand the mcnntngs of their combinations. The internal histogram can partly make it clear, but in practical experiments, we can not often obtain enough length of data to estimate the histogram. Moreover, interval statistics do not directly give us an information about the mechanism of spike event generation. In the present study, we attempt to interpret the combination of inter-spike interval statistics in comparison with a simple stochastic process designed to describe spike events. We define the Markov switching Poisson process, where the state switches in Markov manner between two Poisson processes (one is active state, and the other inactive) at each spike event. Through the Markov switching Poisson process, we interpret the differences in interval statistics of the biological spiking data between middle temporal (MT) area and prefrontal (PF) area of monkey cortex. Most MT data are found to be interpreted as spike sequences whose balance of staying time is biased to inactive state. It is also found that the mean staying time relative to the mean inter-spike interval is shorter than that of typical PF data. The staying time scale can be considered as time scale of temporal correlation in the incoming synoptic inputs. This implies that the differences between the two area originates in the time scales of synaptic inputs correlations.
Yutaka Sakai Shuji Yoshizawa Hiroshi Ohno
Department, of Information and Computer Science,Faculty of Engineering. Saitama University,Saitama 3 Brain Science Research Center,Tamagawa University Research Institute.Machida. Tokyo 194-8610. JAPAN
国际会议
8th International Conference on Neural Information Processing(ICONIP 2001)(第八届国际神经信息处理大会)
上海
英文
935-940
2001-11-14（万方平台首次上网日期，不代表论文的发表时间）