.png){kind=icon}
.png){kind=icon}
Return
The ‘neat’ and ‘messy’ in task-dependent neural geometry and computation
.png){kind=icon}
PRE
AI
C
G
DOI：10.1016/j.tins.2026.04.012.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Advanced analyses of neuronal population activity, along with comparisons to artificial neural networks, have transformed our understanding of how the brain performs and switches between multiple cognitive tasks. Many ‘neat’ features of the brain’s computations have been identified: neural populations encode task-relevant information in a structured way, supporting efficient task switching. However, ‘messy’ features have also been observed: neural activity is modulated by task-irrelevant variables, representations often defy simple normative interpretation, and behavior shows suboptimalities such as switch costs. We propose two paths forward: (i) pursue experiments, analyses, and models that better capture brain-specific constraints and (ii) generate and test hypotheses from artificial network models directly constrained by behavioral and neural data.
Keywords:
neuronal population activity
task-dependent neural geometry
artificial neural networks
cognitive tasks
neural computation
Journal
IF:
15.1
Papers:
5.3K
Citations:
2.2W
Organization
Cited Papers