EVENTS & CALENDAR

MIT SEA GRANT & CENTER FOR OCEAN ENGINEERING SEMINAR

Professor Maarja Kruusmaa

Centre for Biorobotics, Tallinn University of Technology, Estonia

Center for Autonomous Marine Operations and Systems, Norwegian University of Science and Technology, Norway

Longhuan Zhu (PhD Candidate, University of Maine)

Advisors: Dr. Kimberly Huguenard (University of Maine) and Dr. David W. Fredriksson (U.S. Naval Academy)

World aquaculture production has doubled in the past decade and is expected to surpass the capture production this year. Consequently, aquaculture is becoming the dominant source for seafood and other sea products. The rapid expansion of estuarine and nearshore aquaculture will significantly influence the estuarine and coastal ecology, environment and communities. In this presentation, the efforts of researchers from the University of Maine and a network of collaborators are working toward building a Sustainable and Ecological Aquaculture NETwork (SEANET) in Maine will be introduced. Particularly, a 3D computational model for the dynamics of longline aquaculture systems will be demonstrated. This model is being implemented into a computational fluid dynamics (CFD) platform OpenFOAM to model the influence of aquaculture farms on the flow. The lab experiments supported by Nepf Lab (MIT) will also be presented to show the potential of kelp farms on wave attenuation.

Mengjia Xu, PhD

Postdoctoral Associate, MIT McGovern Institute and MGH

In real world applications, networks can be used as a powerful language for describing and modeling complex systems, e.g. social networks, economic networks, biomedical networks, internet networks, etc. In the past few decades, machine learning based graph representation has become an active research area since it can provide insights into how to make good use of the information hidden in graphs and facilitate more accurate and efficient downstream graph analysis tasks (e.g., community detection or link prediction). I will first give an overview of the latest graph embedding methods, and then I will specifically introduce the principles of Gaussian embedding technique with uncertainty quantification. Finally, I will illustrate a real-world functional brain network application using the Gaussian embedding method.

Mengjia Xu is currently a postdoctoral associate at the MIT McGovern Institute and Massachusetts General Hospital. Before joining MGH in 2018, she was a postdoc of Beijing International Center for Mathematical Research at Peking University. She received her PhD degree in Computer Science from Northeastern University of China in 2017. During her PhD, she worked at Brown University for two years as visiting PhD student in the Division of Applied Mathematics. Previously, she worked as a full-time software engineer intern at Neusoft for two years. Her main research interests are to develop data-driven machine learning methods for medical image data analysis in diverse real-world applications.

Current MIT Sea Grant projects in machine learning and modeling:

• Bayesian Intelligent Ocean Modeling and Acidification Prediction Systems (Pierre Lermusiaux)
• Multi-fidelity Modeling for non-linear and multi-valued correlations (Chrys Chryssostomidis)
• Quantifying Acidification Levels in the US North East Coast via Machine Learning Tools (Themistoklis Sapsis)
• The Robotic Intelligent Towing Tank for Learning Complex Fluid-Structure Dynamics (Dixia Fan) >> Read the article