2:40 PM - 3:00 PM
[2L2-OS-6a-05] Reward learning through interaction between self-organization of a spiking neural network and a bodily behavior
Keywords:Self-organization, Spiking neural network, Embodiment
Interaction between the brain network and the body is important for behavioral and cognitive development. Behaviors cause sensory input to the brain, and the neural network self-organizes under the sensory stimulus, which leads to the acquisition of new behaviors. We aim to figure out how this brain-body interaction develops from a viewpoint of a constructive approach. We construct a spiking neural network (SNN) model for the reward learning of canonical babbling, i.e., combination of a vowel and consonant. Vocalization is created by the output of the SNN and is fed back into the SNN. Synaptic weights in the SNN are modulated using STDP. Our experiment showed that STDP under auditory feedback enabled the model to rapidly acquire babbling. We found that neurons in the SNN of the model activated much more when vocalizing a consonant. This body representation in the SNN seems to accelerate the acquisition of babbling.