eduzhai > Applied Sciences > Engineering >

Deep Reinforcement Learning for Neural Control

  • Save

... pages left unread,continue reading

Document pages: 11 pages

Abstract: We present a novel methodology for control of neural circuits based on deepreinforcement learning. Our approach achieves aimed behavior by generatingexternal continuous stimulation of existing neural circuits (neuromodulationcontrol) or modulations of neural circuits architecture (connectome control).Both forms of control are challenging due to nonlinear and recurrent complexityof neural activity. To infer candidate control policies, our approach mapsneural circuits and their connectome into a grid-world like setting and infersthe actions needed to achieve aimed behavior. The actions are inferred byadaptation of deep Q-learning methods known for their robust performance innavigating grid-worlds. We apply our approach to the model of textit{C.elegans} which simulates the full somatic nervous system with muscles and body.Our framework successfully infers neuropeptidic currents and synapticarchitectures for control of chemotaxis. Our findings are consistent with invivo measurements and provide additional insights into neural control ofchemotaxis. We further demonstrate the generality and scalability of ourmethods by inferring chemotactic neural circuits from scratch.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×