Amount Awarded: $20,000
The main purpose of this project is to develop fast control systems using deep neural networks that can be implemented in low-cost embedded devices for decision-making in the timescale of sub-microseconds. In particular, we focus on developing such tools for safety-critical biomedical applications with fast dynamics. Large labeled training data sets are key enablers of deep learning techniques and has a major impact on the accuracy of the trained model. In fact, for some applications, creating labeled training sets is known to be the most time-consuming and expensive part of applying machine learning. This project aims to address this issue by efficiently generating large training data that can be used to develop fast optimization-based control systems. The developed methods will be experimentally validated using biomedical applications of atmospheric pressure plasma jets.