2023 212 E2

On-Device Machine Learning

By: Benediction Simeons

Department: Engineering 

Faculty Advisors: Dr. Zhuwei Qin, Dr. Tom Holton 

This project seeks to implement machine learning model training on a low-power hardware devices (e.g., microcontrollers) that lack rigorous computational power and cloud data direct and/or quick access. This approach of loading ML models on low-power devices is commonly known as TinyML. Specially, I will optimize and deploy one of the most representee machine learning or deep learning model called convolutional neural network (CNN) for the image classification task on the Sony spresense microcontroller. To achieve the on-device training, the neural network architecture will be optimized for efficient resource access by neural network compression.