Rise and subsequent fall in neuro-behavioral coupling during learning a skilled reaching task is revealed by generative AI
- 1Canadian Centre for Behavioural Neuroscience, University of Lethbridge, Lethbridge, AB Canada
Abstract
Understanding complex relations between neuronal activity and animal behavior is central question in neuroscience. Rapid advancements in Artificial Intelligence (AI) methods offer powerful tools to investigate highly non-linear mapping between motor cortex activity and body movements. Here, we developed a Generative Adversarial Network (GAN) that showed that detailed videos of behaving rats can be recreated from activity of just few selected neurons. This analysis also revealed that the predictability of behavior from neuronal activity (and vice versa) initially increases as a rat learns a new task. However, after the animal performance on the motor task achieves the required accuracy, then coupling between neuronal activity and behavior decreases, without degrading task performance. A plausible interpretation is that, as training progresses from Early to Mid training days, more neurons become engaged, forming a denser, broadly distributed representation, which then in the Late training days evolves into a sparse and more energy-efficient representation, with only a small subset of tuned neurons. Neuronal network simulations showed that such changes in coding strategy may be explained by neurons minimizing their energy use. Thus, our approach reveals a non-linear relationship between learning stages and neural-behavioral coupling, which is likely driven by energy efficiency.
Related
- Perceptual lengthening of visual time throughout learning of hand-reaching actions
- Dream incorporation and subsequent memory change after learning and waking reactivation
- Predicting the immediate and subsequent effects of commercials on product valuation using EEG and deep learning
- EEG-based deep learning estimates preterm infants’ cerebral age: Contribution of slow-wave and cross-frequency coupling
- KoopCell: Koopman-Based Generative Model for Learning Single-Cell Dynamics from Distribution Snapshots
BibTeX
@article{tanabe2026rise,
title = {{Rise and subsequent fall in neuro-behavioral coupling during learning a skilled reaching task is revealed by generative AI}},
author = {Sean Tanabe and Michael Eckert and Hardeep Ryait and Ian Q. Whishaw and Masami Tatsuno and Artur Luczak},
journal = {Communications Biology},
year = {2026},
month = jul,
doi = {10.1038/s42003-026-10600-6},
url = {https://doi.org/10.1038/s42003-026-10600-6},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - Rise and subsequent fall in neuro-behavioral coupling during learning a skilled reaching task is revealed by generative AI
AU - Sean Tanabe
AU - Michael Eckert
AU - Hardeep Ryait
AU - Ian Q. Whishaw
AU - Masami Tatsuno
AU - Artur Luczak
PY - 2026
DA - 2026/07/02
JO - Communications Biology
DO - 10.1038/s42003-026-10600-6
UR - https://doi.org/10.1038/s42003-026-10600-6
AB - Understanding complex relations between neuronal activity and animal behavior is central question in neuroscience. Rapid advancements in Artificial Intelligence (AI) methods offer powerful tools to investigate highly non-linear mapping between motor cortex activity and body movements. Here, we developed a Generative Adversarial Network (GAN) that showed that detailed videos of behaving rats can be recreated from activity of just few selected neurons. This analysis also revealed that the predictability of behavior from neuronal activity (and vice versa) initially increases as a rat learns a new task. However, after the animal performance on the motor task achieves the required accuracy, then coupling between neuronal activity and behavior decreases, without degrading task performance. A plausible interpretation is that, as training progresses from Early to Mid training days, more neurons become engaged, forming a denser, broadly distributed representation, which then in the Late training days evolves into a sparse and more energy-efficient representation, with only a small subset of tuned neurons. Neuronal network simulations showed that such changes in coding strategy may be explained by neurons minimizing their energy use. Thus, our approach reveals a non-linear relationship between learning stages and neural-behavioral coupling, which is likely driven by energy efficiency.
SN - 2399-3642
ER -