정보통신대학원 세미나를 소개합니다.
이번 학기는 온라인 or 오프라인으로 진행 되며, 학부생, 대학원생, 교직원 누구나 참여 가능합니다.
▶▶세미나1-정보통신대학원 세미나 2021년 2학기
Promising Collaboration with Neuroscience and Computer Science: Neuro-Information Processing Underlying Reinforcement Learning
홍사익 교수 (한동대 정보통신연구소)
Recently, neuroscience and computer science drive each other forwards. However, there are still many gaps in understanding the two disciplines and this hinders better academic achievement. Here, this seminar will elucidate 1) behavioral tasks for reinforcement learning, 2) brain circuits related to model-based and mode-free reinforcement learnings, and 3) application of computer science to neuroscience. Reinforcement learning aims to increase or reduce the rate of an agent’s action based on the reward or punishment received from an environment. Effort-based reward delivery shapes goal-directed behavior with flexible action strategy, whereas time-based reward delivery forms habit with fixed action selection. Maladaptive and uncontrollable habits are the hallmark of mental disorders, such as obsessive-compulsive disorders and addiction. A cortico-striatopallidal-thalamocortical loop is the key brain circuit for reinforcement learning. Especially, basal ganglia harboring inhibitory striatopallidal circuits and star-shaped astrocytes codifies action policy. Interestingly, temporal dynamics of astrocytes in the basal ganglia are silenced during habit learning. In addition, manipulation of astrocytes in the basal ganglia restores model-based performances after model-free reinforcement learning. Moreover, computer neuroscientific approaches, such as support vector machine and tensor component analysis, support the promising roles of astrocytes in reinforcement learning and provide insight into a potential therapeutic target for mental disorders.
＊일시 9월 8일 (수), 13시
＊강의실 NTH412호
＊회의 ID 862 4495 4271
＊문의 4단계 BK21 AI사업단 (260-3150)
