The primary objective of these studies was to investigate motor learning processes through innovative approaches: (i) a new combined take that integrates motor sequence learning and motor adaptation; (ii) developing a computational model able to mimic real healthy subjects data. In this work, we first investigated the interaction of visuomotor adaptation and sequence learning in the early acquisition phase. Our results showed that simultaneous learning selectively affects both in different ways compared to sequence learning and rotation adaptation alone. Furthermore, our results showed significant differences in the number of anticipatory movements by comparing the combined task with the sequence learning alone. In particular, at the end of the training phase, subjects better anticipate the sequence in the explicit task alone. This finding corroborates a higher verbal score in the explicit task than combined. A second study aimed to understand motor learning processes in their early stage by computing a computational model that mimics the neurophysiological mechanism underpinning motor learning. We developed an integrated model (characterized by a combination of supervised and reinforcement learning) that exhibits a faster learning process concerning the classic reinforcement learning model. This faster learning process of the integrated model was due to its improved exploration strategy. In particular, since the model gets the reward for the first time, the supervised component can learn a first coarse trajectory to reach the target. This trajectory affects the exploration by reducing its randomness and making it more focused on the target. The RL component then gradually improves the sub-optimal initial solution trial after trial while avoiding unneeded explorations as in the pure RL model. Similarly, the SL component gradually improves its behavior by receiving a better teaching signal (trajectories) from the RL processes. This mutual training between SL and RL thus produced a faster learning process.

Motor learning in Parkinson's Disease

PONTE, CHIARA
2022-08-30

Abstract

The primary objective of these studies was to investigate motor learning processes through innovative approaches: (i) a new combined take that integrates motor sequence learning and motor adaptation; (ii) developing a computational model able to mimic real healthy subjects data. In this work, we first investigated the interaction of visuomotor adaptation and sequence learning in the early acquisition phase. Our results showed that simultaneous learning selectively affects both in different ways compared to sequence learning and rotation adaptation alone. Furthermore, our results showed significant differences in the number of anticipatory movements by comparing the combined task with the sequence learning alone. In particular, at the end of the training phase, subjects better anticipate the sequence in the explicit task alone. This finding corroborates a higher verbal score in the explicit task than combined. A second study aimed to understand motor learning processes in their early stage by computing a computational model that mimics the neurophysiological mechanism underpinning motor learning. We developed an integrated model (characterized by a combination of supervised and reinforcement learning) that exhibits a faster learning process concerning the classic reinforcement learning model. This faster learning process of the integrated model was due to its improved exploration strategy. In particular, since the model gets the reward for the first time, the supervised component can learn a first coarse trajectory to reach the target. This trajectory affects the exploration by reducing its randomness and making it more focused on the target. The RL component then gradually improves the sub-optimal initial solution trial after trial while avoiding unneeded explorations as in the pure RL model. Similarly, the SL component gradually improves its behavior by receiving a better teaching signal (trajectories) from the RL processes. This mutual training between SL and RL thus produced a faster learning process.
Parkinson's Disease
Motor Learning
Artificial Intelligence
File in questo prodotto:
File Dimensione Formato  
phdunige_3710863.pdf

embargo fino al 30/08/2023

Descrizione: final document
Tipologia: Tesi di dottorato
Dimensione 9.57 MB
Formato Adobe PDF
9.57 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11567/1094283
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact