Intrinsic Behavioral Variability Facilitates Flexible Representations: A Neuromotor Developmental Perspective
Keywords: motor, development, adaptation, simulation, supervised-learning, unsupervised-learning
Abstract: Dynamic human movement necessitates a dynamic representation of the body. The mechanisms underlying the initiation, development, and maintenance of such representations can provide a biological perspective to developing more flexible representations within computational agents. Taking inspiration from the prenatal twitches shown to initiate the human neuromotor representation, we question how these same twitches, present throughout development, may also facilitate subsequent motor adaptation. Across three experiments, we examine the influence twitches, as a form of intrinsic behavioral variability, may have in facilitating motor adaptation to novel situations. In a series of simulated reaching tasks, we trained agents to reach targets while overcoming behavioral, physiological, and neurological changes. Overall, we found evidence that agents exposed to intermittent behavioral variability outperformed their counterparts, showing greater neural weight variability, indicative of greater exploration. Taken together, this work provides a biologically plausible computational framework for flexible representation development.
Primary Area: applications to neuroscience & cognitive science
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 1936
Loading