Dynamics of Learning: Lessons from Living Systems to Artificial Neural Networks

Abstract:
What is learning, and how does it emerge in both living systems and machines? In this talk, we first introduce a concise and general definition: learning is the process by which a system acquires information, encodes and stores it in its internal degrees of freedom (memory), and uses this memory to improve performance on specific tasks. This perspective naturally casts learning as a dynamical process evolving in time. Within this theoretical framework, we compare learning mechanisms across a broad spectrum of systems—from adaptive responses in single cells, to associative learning in biological neural networks, to algorithmic methods such as stochastic gradient descent in artificial neural networks.