Co-hosted an organized session "OS-22 Simulation and AI" with Osaka University, National Institute of Advanced Industrial Science and Technology, NEC, and others at the 2022 National Conference of the Japanese Society for Artificial Intelligence

In this session, we will present research on using machine learning for simulation and research on incorporating simulation into machine learning, which has been actively undertaken in recent years for the real-world application of artificial intelligence technology, and will discuss the essential aspects of simulation and machine learning. Explore the possibilities of fusion.

Traditionally, simulation using deductive calculations and AI/machine learning using inductive calculations have had opposing research directions. There are many types of simulation, including finite element method, molecular dynamics simulation, and multi-agent simulation, and there are a wide variety of real-world applications such as structure design, new material development, and traffic/people flow control. On the other hand, in the field of machine learning, with the development of surrogate model technology using deep learning and Gaussian processes, it has become possible to approximate real-world systems with high accuracy as a black box. In recent years, new research has been expanding in each field by utilizing each other's methods and models.

In the simulation field, we are using surrogate models that replace part of the simulator with machine learning models, and data assimilation that statistically optimizes simulation parameters from measured data using approximate Bayesian calculations (ABC), Kalman filters, etc. By introducing it, calculation speed and accuracy have been increased. On the other hand, in the field of machine learning, new learning problems are discovered and approximate solutions are developed by embedding first-principles calculations and simulators in parts of models. However, there is a tendency for issues to be set depending on the domain in which simulation is applied, and the results may be limited to developments in individual fields. In response to the above situation, this session will explore the potential of essential fusion of simulation and machine learning.


The event will be held in a hybrid format between the Kyoto International Conference Center and online, and will be held on June 14, 2022 from 14:20 to 17:00, so please join us. For details on the presentation program, etc., please see the homepage of this session.
Frame 1)
Frame 2)
Please refer to the.

BIRD INITIATIVE Co., Ltd. will continue to release AI-related technology and give back to society.

Latest news