A new paradigm for perceptual learning in neocortical dendrites and deep neural networks

新皮质树突和深度神经网络感知学习的新范式

• 批准号: RGPIN-2022-05033
• 负责人: Muller, Eilif
• 金额: $ 2.77万
• 依托单位: Université de Montréal
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=761471
• 关键词: new; paradigm; perceptual; learning; neocortical

The fields of neuroscience and artificial intelligence have a long history of fertile bi-directional interactions. On the one hand, important inspiration for the development of artificial intelligence systems has come from the study of natural systems of intelligence, the mammalian neocortex in particular. On the other, important inspiration for models and theories of the brain have emerged from artificial intelligence research. A central question at the intersection of these two disciplines is how sensory perception is learned in the neocortex. In the field of deep learning, neocortical learning inspires active areas of research aiming at matching its data efficient incremental learning and out-of-distribution generalization, but largely ignoring decades of neuroscientific insights or constraints. Building on recent advances in our understanding of the synaptic, dendritic and circuit physiology of the neocortex, the ABL-Lab will leverage a synergy of biophysical simulations of neocortical circuits and functional deep neural networks to explore a radically new perspective on the processes of learning in the neocortex, providing insight into the function and dysfunction of neocortical learning, and also new inspiration to solve fundamental challenges for deep learning.

神经科学和人工智能领域有着丰富的双向互动的悠久历史。一方面，人工智能系统发展的重要灵感来自于对自然智能系统的研究，特别是对哺乳动物新皮层的研究。另一方面，人工智能研究为大脑模型和理论提供了重要的灵感。这两个学科交叉的一个核心问题是，感觉知觉是如何在新皮层中习得的。在深度学习领域，新皮层学习激发了活跃的研究领域，旨在匹配其数据高效增量学习和分布外泛化，但在很大程度上忽略了几十年来神经科学的见解或限制。基于我们对新皮层的突触、树突和回路生理学的最新理解，abl实验室将利用新皮层回路和功能性深度神经网络的生物物理模拟的协同作用，探索新皮层学习过程的全新视角，为新皮层学习的功能和功能障碍提供见解，并为解决深度学习的基本挑战提供新的灵感。