The verification's criterion of learning algorithm

摘要

Construction of an effective real time learning algorithm is mostly needed for the pulsed neural network nowadays. The learning algorithm should collect and convert the data from networks inputs into a PNN memory while network's working. Keeping of an asynchronous state in the background is a big challenge during learning PNN with a certain number of recurrent connections like Hebbian cell assemblies (HCA) or synfire chains (SFC). This paper presents methods which are usable for refusing or accepting the examined learning algorithm in the relatively short time of simulation and it gives us advice about direction of our research.