Temporal Hebbian Plasticity Designed for Efficient Competitive Learning

摘要

Understanding the functional roles of temporal Hebbian plasticity has been of growing interest since several experiments revealed that the change in synaptic efficacy was determined by the precise temporal relation between post- and presynaptic spikes. We here investigate the learning properties of the typical synaptic modification forms. We explain how the peculiar characteristics in synaptic modification, such as asymmetry, decay rate, and oscillatory behavior, exert effects on the direction and the performance of network formation. Also, we argue that the aforementioned characteristics help to achieve proper network adaptation, such as activity-dependent columnar organization, through an efficient competitive learning process.