Contributed talk in Philosophy, Language, Art & Education 3, Aug. 2, 2019, 5 p.m. in room USB.G.003

Spatial Representation of Self and Other by Superposed Neural Network Model

Wataru Noguchi, Hiroyuki Iizuka, Shigeru Taguchi, Masahito Yamamoto

watch Publication We propose a neural network model that develops shared representation of self and other's location based on recent findings of hippocampal cells that represent self and other's location without distinction. Our proposed network is constructed to have a superposed states of the self- and other-related representation; two modules with the same network weights process different inputs at the same time. The proposed network was trained to predict subjective vision of the self agent in a simulated environment where other agent exists. As a results of the training, shared representation of self and other's location in our proposed network model was self-organized like in our brain. We considered that having superposed states for self and other is required to develop the shared representation of self and other's location that can realize the association between self and other.