Core Concepts
Adaptive deep brain stimulation using ϵ-Neural Thompson Sampling improves treatment efficacy and energy efficiency in Parkinson's disease.
Abstract
Deep Brain Stimulation (DBS) is effective for Parkinson's disease motor symptoms.
Traditional DBS devices have limitations, leading to research on adaptive DBS (aDBS).
Contextual Multi-Armed Bandit (CMAB) approach proposed for aDBS.
ϵ-Neural Thompson Sampling algorithm balances exploration and exploitation effectively.
Evaluation using computational Basal Ganglia Model shows improved performance over existing methods.
Stats
RL approaches require significant training data and resources.
CMAB leads to better sample efficiency compared to RL.
ϵ-NeuralTS outperforms existing cDBS methods and baselines.