Utilizing Machine Learning to Predict Variable Threshold Target Speed for Future Motor Control Studies of the Brain
Open Access
Author:
Muthler, Noah
Area of Honors:
Electrical Engineering
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
Tarkeshwar Singh, Thesis Supervisor Julio Urbina, Thesis Honors Advisor
Keywords:
Machine Learning Logistic Regression Supervised Learning Augmented Reality Neuroscience Motor Control
Abstract:
A successful study involving interception planning of the brain and execution of the motor control system depends highly on the simulation’s target speed, proven by Fitts’ law. Currently, there lacks a way to efficiently and effectively determine a target speed that results in a desired accuracy rate. This study provides a supervised learning logistic regression model that proves to expedite the process of determining a variable threshold target speed. This algorithm consists of two decoupled steps: first, data exported from a simple middle out reaching task trains the model, and second, the model predicts a variable output threshold speed to be used in a more complex subsequent motor control study. The algorithm will normalize the input data, utilize gradient descent with momentum to tune the model, and reduce variance with cross-validation.