The computational foundations of dynamic coding in working memory

The prefrontal cortex is a key region involved in working memory.Working memory related neural activities in the prefrontal cortex exhibit unexpectedly rich and complex dynamics during even the simplest tasks — a phenomenon called ‘dynamic coding'.Task-optimized neural networks also exhibit dyn...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Trends in cognitive sciences 2024-07, Vol.28 (7), p.614-627
Hauptverfasser: Stroud, Jake P., Duncan, John, Lengyel, Máté
Format: Artikel
Sprache:eng
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The prefrontal cortex is a key region involved in working memory.Working memory related neural activities in the prefrontal cortex exhibit unexpectedly rich and complex dynamics during even the simplest tasks — a phenomenon called ‘dynamic coding'.Task-optimized neural networks also exhibit dynamic coding when optimized on working memory tasks.Two key aspects of a neural network determine whether it exhibits dynamic coding: the connectivity of the network and the inputs it receives.Dynamic coding results from an optimality principle of robust working memory maintenance. Working memory (WM) is a fundamental aspect of cognition. WM maintenance is classically thought to rely on stable patterns of neural activities. However, recent evidence shows that neural population activities during WM maintenance undergo dynamic variations before settling into a stable pattern. Although this has been difficult to explain theoretically, neural network models optimized for WM typically also exhibit such dynamics. Here, we examine stable versus dynamic coding in neural data, classical models, and task-optimized networks. We review principled mathematical reasons for why classical models do not, while task-optimized models naturally do exhibit dynamic coding. We suggest an update to our understanding of WM maintenance, in which dynamic coding is a fundamental computational feature rather than an epiphenomenon. Working memory (WM) is a fundamental aspect of cognition. WM maintenance is classically thought to rely on stable patterns of neural activities. However, recent evidence shows that neural population activities during WM maintenance undergo dynamic variations before settling into a stable pattern. Although this has been difficult to explain theoretically, neural network models optimized for WM typically also exhibit such dynamics. Here, we examine stable versus dynamic coding in neural data, classical models, and task-optimized networks. We review principled mathematical reasons for why classical models do not, while task-optimized models naturally do exhibit dynamic coding. We suggest an update to our understanding of WM maintenance, in which dynamic coding is a fundamental computational feature rather than an epiphenomenon.
ISSN:1364-6613
1879-307X
1879-307X
DOI:10.1016/j.tics.2024.02.011