Intelligence of Artificial Intelligence : A Study on Brain Decoding using Machine Learning and the Analysis of Brain Functions

In this paper, we propose a brain decoding method for the generation of natural language descriptions based on human brain activity by a deep learning model. In addition, to elucidate brain functions, we focused on predictive coding, which is hypothesized to be one of the functions of the cerebral cortex. We investigated the agreement between the predictions of deep learning model and the actual brain activity data to test the predictive coding hypothesis. Furthermore, we propose a method for generating images from human brain activity data by using the same validation model. (Received November 9, 2020;accepted December 7, 2020)