Autonomous Regulation of Self and Non-Self in Embodied Neural Networks.

Abstract

Our previous studies has shown that embodied cultured neural networks and spiking neural networks with spike-timing dependent plasticity (STDP) can be learned as if it avoids stimulation from outside. In a sense, embodied neural networks can autonomously change their activity to avoid an external stimuli. It has known that there are two ways to learn a behavior to avoid the stimulation for the networks: (1) reinforcement of a behavior which can avoid external stimuli, (2) suppressing of a behavior which induces external stimuli. We found there is also a third property for avoiding stimulation where if avoiding external stimuli is difficult, then, network decrease stimulus-evoked spikes as if it ignores input neurons. We also show these results are reproduced by spiking neural networks with asymmetric STDP. These properties can be regarded as autonomous regulation of self and non-self for the network.