Computation noise promotes zero-shot adaptation to uncertainty during decision-making in artificial neural networks
Science Advances · 2024-11 · 10 (44)
DOI: 10.1126/sciadv.adl3931 ↗Kurz gefasst
Fachartikel in der Zeitschrift Science Advances, 2024-11. Verfasst von Charles Findling und 1 weiteren Autor:innen. Beteiligte Einrichtung: Laboratoire de Neurosciences Cognitives et Computationnelles, Institut National de la Santé et de la Recherche Médicale (Inserm), Paris, France. Hauptthemen laut MEDLINE: Neural Networks, Computer, Decision Making. Die Arbeit wurde bisher 17-mal zitiert (Crossref) und nennt 57 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar.
Abstract (Originalsprache)
Random noise in information processing systems is widely seen as detrimental to function. But despite the large trial-to-trial variability of neural activity, humans show a remarkable adaptability to conditions with uncertainty during goal-directed behavior. The origin of this cognitive ability, constitutive of general intelligence, remains elusive. Here, we show that moderate levels of computation noise in artificial neural networks promote zero-shot generalization for decision-making under uncertainty. Unlike networks featuring noise-free computations, but like human participants tested on similar decision problems (ranging from probabilistic reasoning to reversal learning), noisy networks exhibit behavioral hallmarks of optimal inference in uncertain conditions entirely unseen during training. Computation noise enables this cognitive ability jointly through “structural” regularization of network weights during training and “functional” regularization by shaping the stochastic dynamics of network activity after training. Together, these findings indicate that human cognition may ride on neural variability to support adaptive decisions under uncertainty without extensive experience or engineered sophistication.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- Laboratoire de Neurosciences Cognitives et Computationnelles, Institut National de la Santé et de la Recherche Médicale (Inserm), Paris, France.
- Département des Neurosciences Fondamentales, Université de Genève, Geneva, Switzerland.
- Département d'Études Cognitives, École Normale Supérieure, Université PSL, Paris, France.
- Institut du Psychotraumatisme de l'Enfant et de l'Adolescent, Conseil Départemental Yvelines et Hauts-de-Seine, Versailles, France.
ORCID-Profile: Charles Findling, Valentin Wyart
Zitieren
APA
Findling, C., & Wyart, V. (2024). Computation noise promotes zero-shot adaptation to uncertainty during decision-making in artificial neural networks. Science Advances, 10(44). https://doi.org/10.1126/sciadv.adl3931
Vancouver
Findling C, Wyart V. Computation noise promotes zero-shot adaptation to uncertainty during decision-making in artificial neural networks. Science Advances. 2024;10(44). doi:10.1126/sciadv.adl3931
BibTeX
@article{Findling2024,
author = {Findling, Charles and Wyart, Valentin},
title = {Computation noise promotes zero-shot adaptation to uncertainty during decision-making in artificial neural networks},
journal = {Science Advances},
year = {2024},
volume = {10},
number = {44},
doi = {10.1126/sciadv.adl3931}
}
EUVIMED