Announcement of Public Defense
July 11, 2025
12 – 1:30pm
HUMAN ACTIVITY RECOGNITION AND IDENTIFICATION DRIVEN AUTOMATED DEEP LEARNING FOR TIME-SERIES CLASSIFICATION
Justin Alan Gamble
B.S May 2009, Louisiana Tech 91¶ÌÊÓƵ
M.E August 2017, Old Dominion 91¶ÌÊÓƵ
Director: Holly A. H. Handley, PhD, PE
Ìý
ABSTRACT
Designing and optimizing machine learning models, particularly complex deep learning architectures, remains a significant challenge due to the vast search space of configurations and the intricate dependencies among hyperparameters and network structures. To address this challenge and make deep learning experimentation more accessible to a wider audience, this dissertation investigates the use of Automated Machine Learning (AutoML) tailored to the time series classification tasks of human activity recognition and identification. However, manually tuning neural architectures and their hyperparameters can be inefficient, inconsistent, and time-consuming. This study implements a neural architecture search (NAS) framework, comparing traditional methods of grid and random search with advanced AutoML techniques employing Optuna and modified gamma search strategy. More than 800,000 convolutional neural network (CNN) architectures were evaluated using the MotionSense dataset. The results demonstrate that automated search methods consistently outperformed manual tuning, achieving higher F1 scores and greater computational efficiency. Collectively, this research provides an empirical foundation, reusable datasets, and a practical AutoML framework that together lower the barriers to deep learning experimentation. The systematic architecture exploration delivers a high-performing participant identification model based on behavioral signatures, supporting more accessible, efficient, and scalable AI integration in digital engineering contexts, with relevance for rapid, window-based identification and continuous authentication applications.
Ìý
Zoom Meeting Information:
Meeting ID: 988 2351 7993
Passcode: 940543
Ìý
Biography
Justin Gamble earned M.E. in Systems Engineering from Old Dominion 91¶ÌÊÓƵ in 2017 and a B.S. in Electrical Engineering from Louisiana Tech 91¶ÌÊÓƵ in 2009. He is the Lead Software Engineer for the Special Technology Integration Branch at Naval Surface Warfare Center Dahlgren Division (NSWCDD).