Aurelien Meray

Ph.D. Candidate


Lecture Information:
  • June 14, 2023
  • 10:00 AM
  • Zoom
Ph.D. Candidate Aurelin Meray's Headshot

Speaker Bio

Aurelien Meray is a Ph.D. candidate at the Knight Foundation School of Computing and Information Sciences (KFSCIS) at Florida International University (FIU). With a background in Python, machine learning, and data analysis, Aurelien has contributed to various environmental monitoring projects. As a Graduate Researcher at the Applied Research Center, FIU, Aurelien led a 2-year data analysis and machine learning project, collaborating with a Nuclear Engineering Professor at MIT, and presenting research at the Waste Management Symposia Conference. He also interned at Lawrence Berkeley National Laboratory, where he developed a machine learning package for soil and groundwater data analysis and implemented a sensor placement optimization algorithm for the Advanced Long-Term Monitoring Systems (ALTEMIS) project. Aurelien further honed his skills as an Intern at Frontier Development Lab, where he led the implementation of preprocessing and development of a cutting-edge surrogate model using U-FNO and PyTorch for the Climate Adaptation: Digital Twin: Environmental Remediation project. He has experience utilizing GCP, managing large simulation data, and effectively communicating technical ML information to non-technical clients and DOE personnel.

Abstract

Environmental monitoring plays a vital role in understanding and managing our natural environment. With the advent of sensor technologies and the Internet of Things (IoT), large-scale sensor networks have been deployed to collect environmental data. However, deploying and maintaining a large number of sensors can be costly and resource-intensive, necessitating the optimization of sensor selection. This PhD proposal focuses on developing a novel approach for Multivariate Subset Sensor Selection Optimization (SSSO) in environmental monitoring networks. The research objectives include creating a dimensionality reduction method for compressing multivariate and multi-timestep data, formulating the sensor placement optimization problem using the compressed data representation, analyzing various optimization algorithms, and evaluating the proposed method’s performance. The research will make use of two real-world environmental monitoring datasets, the F-Area Historical dataset from the Department of Energy’s Savannah River Site (SRS) F-Area and the temperature field dataset from the Berkeley Intel Lab. By achieving these objectives, this study aims to contribute to the fields of computer science and environmental monitoring by providing a comprehensive framework for optimizing sensor selection in environmental monitoring networks.