AUTHOR=Osuna-Orozco Rodrigo , Zhao Yi , Stealey Hannah Marie , Lu Hung-Yun , Contreras-Hernandez Enrique , Santacruz Samantha Rose 

TITLE=Adaptation and learning as strategies to maximize reward in neurofeedback tasks

JOURNAL=Frontiers in Human Neuroscience

VOLUME=Volume 18 - 2024

YEAR=2024

URL=https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2024.1368115

DOI=10.3389/fnhum.2024.1368115

ISSN=1662-5161

ABSTRACT=This study explores the dynamics of neural activity during a center-out reaching task with continuous visual feedback. Results for a brain-computer interface task performed by two nonhuman primate (NHP) subjects are compared to simulations from a reinforcement learning agent performing an analogous task. Principal component analysis (PCA) of the spiking activity reveals that the neural manifold is largely preserved during a session where a rotational perturbation is introduced. Adaptation of neural activity occurs within said manifold and rotations are compensated by reassignment of regions of the neural space in an angular pattern that cancels said rotations. The artificial neural network encoding the policy of the reinforcement learning agent generates activity that also lies in a low-dimensional manifold that is isomorphic with the task. However, retraining of the agent results in substantial modifications of the underlying manifold. Our findings demonstrate that NHPs adapt their existing neural dynamic repertoire in a quantitatively precise manner to account for perturbations of different magnitudes.