Parsimonious Deep Learning with Human Feedback Under Computational Constraints
Department of Electrical and Computer Engineering
Location: Burchard 714
Speaker: Prof. Nikhil Muralidhar, Department of Computer Science, Stevens Institute of Technology
The recent confluence of deep learning (DL) and the internet-of-things (IoT) has brought with it a unique set of application challenges and opportunities. While the opportunities comprise in developing DL pipelines for modeling complex processes in critical infrastructure systems (CIS) like communication networks and manufacturing contexts, the main challenge is posed by the computational and memory constraints in these IoT contexts. In this talk we explore strategies to develop computationally parsimonious DL pipelines for tasks like anomaly detection and image segmentation inCIS process monitoring pipelines with the ability to incorporate sparse feedback from human experts.
Nikhil Muralidhar is an Assistant Professor in the CS department at Stevens Institute of Technology. At Stevens, Nikhil directs the Scientific Artificial Intelligence (ScAI) Lab with a research focus on applied machine learning (ML) in domains like physics, fluid dynamics, computational epidemiology, and cyber-physical systems (CPS). One of Nikhil's current research interests in cyber-physical systems is to develop anomaly detection and CPS modeling ML pipelines that are computationally parsimonious and able to learn with sparse human feedback.