题目: A Synaptic Story of Working Memory（工作记忆的突触理论）

时间：2023年6月7日（周三）上午10:00-11:30

地点：心理学院520

主讲人: 弭元元 教授

主讲人简介：

弭元元，重庆大学医学院神经智能研究中心教授。2012年于北京师范大学获得理论物理专业博士学位，并先后在以色列Weizmann Institute of Science和美国Columbia University做博士后。研究方向为计算神经科学。专注于采用数理建模和计算仿真的方法研究脑在网络层面处理动态信息的一般性原理，包括工作记忆的容量与调控、时空信息的网络编码等；并基于此发展了类脑运动模式的快速识别算法、运动目标的预测追踪算法等。合作指导的课题获得《首届全国大学生类脑计算创新应用大赛暨国际邀请赛》总决赛一等奖（2017年）。以第一或通讯（含共同）在神经科学领域刊物Neuron, PNAS, Progress in Neurobiology等，和人工智能领域的国际会议NeurIPS，Neural Networks等，发表论文20余篇。获得国家自然科学基金委交叉学部优秀青年基金，北京市科技新星等项目的支持。

讲座内容：Working memory refers to the phenomenon that the brain can store and manipulate input information temporally, in order to carry out certain cognitive functions. The neural mechanism underlying working memory has been debated constantly in the field. The synapse-based theory proposes that rather than relying on persistent neuronal firings, the temporally facilitated synaptic strengths due to short-term synaptic plasticity (STP) can carry the memory trace of inputs and thus support working memory. In this talk, I will introduce two studies aligning with this STP-based theory. First, we built up a computational model to analyze the capacity of working memory and found that the model naturally predicts the capacity of 4~6, when the biological plausible parameters are used, and the model works in a way analogy to the theta-gamma oscillation. Secondly, we manipulated the synapse strengths via STP by applying external probe signals, and found that the memory retrieval performances predicted by the model agree very well with the experimental findings. Overall, our studies suggest that STP can serve as the neural mechanism for working memory.