WeSCoS Colloquium #12 “Learning energy conservation/dissipation systems based on Gaussian processes” By Yusuke Tanaka (Research Scientist, NTT Communication Science Laboratories, NTT Corporation.)

WeSCoS Colloquium #12 “Learning energy conservation/dissipation systems based on Gaussian processes”
Presenter: Yusuke Tanaka (Research Scientist, NTT Communication Science Laboratories, NTT Corporation)
Date and time: 3:00 pm-, November 25, 2024 (Monday)
Abstract:
Data-driven physical modeling and simulation based on machine learning is attracting attention.
In order to obtain highly accurate machine learning models, data of sufficient quality and quantity is essential, but in reality, incomplete data containing noise and defects are often provided. In such situations, one effective approach is to utilize prior knowledge to compensate for data incompleteness.
In this lecture, we will use the Gaussian process, which is one of the most commonly used methods in the machine learning field, as a basis and integrate the theory of Hamiltonian mechanics to predict with high accuracy physical phenomena that follow energy conservation and dissipation measurements. We will introduce the method.

Venue: Zoom Pre-registration is required. Please register from here by Noon November 21. We will send the Zoom URL on November 21.
·Language: Japanese