What should a neuron aim for? Designing local objective functions based on information theory

作     者：Schneider, Andreas C. Neuhaus, Valentin Ehrlich, David A. Makkeh, Abdullah Ecker, Alexander S. Priesemann, Viola Wibral, Michael 

作者机构：Faculty of Physics Institute for the Dynamics of Complex Systems University of Göttingen Germany Max Planck Institute for Dynamics and Self-Organization Germany Campus Institute for Dynamics of Biological Networks University of Göttingen Germany Institute of Computer Science Campus Institute Data Science University of Göttingen Germany 

出 版 物：《arXiv》 (arXiv)

年 卷 期：2024年

核心收录：

主　　题：Information theory 

摘      要：In modern deep neural networks, the learning dynamics of the individual neurons is often obscure, as the networks are trained via global optimization. Conversely, biological systems build on self-organized, local learning, achieving robustness and efficiency with limited global information. We here show how self-organization between individual artificial neurons can be achieved by designing abstract bio-inspired local learning goals. These goals are parameterized using a recent extension of information theory, Partial Information Decomposition (PID), which decomposes the information that a set of information sources holds about an outcome into unique, redundant and synergistic contributions. Our framework enables neurons to locally shape the integration of information from various input classes, i.e. feedforward, feedback, and lateral, by selecting which of the three inputs should contribute uniquely, redundantly or synergistically to the output. This selection is expressed as a weighted sum of PID terms, which, for a given problem, can be directly derived from intuitive reasoning or via numerical optimization, offering a window into understanding task-relevant local information processing. Achieving neuron-level interpretability while enabling strong performance using local learning, our work advances a principled information-theoretic foundation for local learning strategies. Copyright © 2024, The Authors. All rights reserved.