Investigation of Pair and Triplet-Based Spike-Timing Dependent Plasticity Mechanism for Spiking Context-Dependent Learning

Context-dependent learning is the concept inspired by the observation of the learning mechanism of brain, which endows spiking neuron network with the ability of recognition in the navigation of the concerted context. This paper provides the mechanism investigation of spike-timing dependent plasticity for context-dependent learning based on spiking neural network model. Two types of spike-timing dependent plasticity learning rule are considered, including pair-based and triplet-based spike-timing dependent plasticity. Based on several experiments, parameters critical to the learning efficiency of the context-dependent task are dedicatedly discussed. The amplitude that controls the intensity of potentiation should be much stronger than that of depression for pair-based learning rule. The time constant which decides the step width should match the frequency of spiking. Compared to the pair-based method, triplet-based learning method owns greater sensitivity to frequency and stronger selectivity in context-dependent task. The amplitude contained in the third order provides subordinate advantages as well as side-effect. This study can be further applied in a number of fields, such as unmanned vehicle, edge computing of Internet-of-things, and bio-inspired robotics.