UROP Project

Machine Learning, Internet of Things, Optimization, Tiny ML

Research Mentor: Dr. Marcos Muller Vasconcelos, He/Him/His
Department, College, Affiliation: Electrical and Computer Engineering, FAMU-FSU College of Engineering
Contact Email: m.vasconcelos@fsu.edu
Research Assistant Supervisor (if different from mentor):
Research Assistant Supervisor Email:
Faculty Collaborators:
Faculty Collaborators Email:

Looking for Research Assistants: Yes
Number of Research Assistants: 2
Relevant Majors: Electrical Engineering
Computer Engineering
Computer Science
Mathematics
Statistics
Project Location: FAMU-FSU College of Engineering
Research Assistant Transportation Required: Yes
Remote or In-person: Partially Remote
Approximate Weekly Hours: 5-10 hours a week,
Roundtable Times and Zoom Link: Not participating in the Roundtable

Project Description

The Internet of Things (IoT) -- a complex ecosystem that interconnects smartphones, tablets, machine-type devices, people, and mundane objects into a large-scale network -- has flooded the world with inexpensive devices that collect, store, process and communicate data at unprecedented volumes. Although the cost of sensing and storing data is relatively small, the power required to process and share data is often large and restrictive. Therefore, it is imperative to identify, process, and transmit only the most informative data. The adoption of cleverly designed data selection policies allows concurrent applications to run on limited shared resources, leading in enhanced efficiency and economic gains. In the case of IoT, even modest gains in performance mount quickly due to the massive scale of the system.

The overall goal of this project is to develop a new system architecture for the selection of the most informative data in multi-agent networks based on the novel notion of Maximum Disagreement and its application in distributed machine learning with IoT devices, as well as implementation in a combination of client-server and peer-to-peer networks. The rationale for the work is twofold: First, we will design a new distributed algorithm based on Maximum Disagreement for machine learning that operates in two layers -- a local network and the cloud infrastructure. Secondly, we will implement this new system in a real testbed using TinyML devices under strict communication constraints while ensuring privacy and robustness guarantees. Achieving these objectives could significantly improve performance and convergence speed, particularly in large-scale commercial distributed systems.

Research Tasks: Literature Review
Programming Tiny ML/IoT devices
Computer Simulations
Communication Networks
Data Collection and Analysis


Skills that research assistant(s) may need: Machine Learning
Wireless Communication Networks
Tinkering
Problem-Solving