| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Artificial Intelligence Applications in Integrated Computational Materials Engineering
|
| Presentation Title |
ANNA: An Open-Source Platform for Developing Artificial Neural Networks Assistant Potential Enabling High Accurate and Efficient Molecular Dynamics Simulation |
| Author(s) |
Meng Zhang, Junya Inoue |
| On-Site Speaker (Planned) |
Meng Zhang |
| Abstract Scope |
A novel force field framework, artificial neural network (ANN) assistant potential, was proposed by incorporating an ANN into typical potential functions. The resulting potential for body-centered cubic (BCC) iron demonstrates exceptional accuracy in property predictions compared with ANN and traditional potentials while boasting remarkable computational efficiency. Its performance overcomes both 12-MPI CPU-only ANN potential and GPU-accelerated ANN potential by achieving speedups of 201x and 26x, respectively, which is very close to the performance of traditional potentials. The proposed approach has a potential to provide a powerful way to develop high-performance and high-accuracy potentials even in the other systems. However, training the ANN-assistant potential is more complex compared to traditional ML potentials. To address this challenge, an open-source package, named ANNA, was developed for training these potentials. The construction and application of ANN assistant potentials using ANNA is demonstrated through an example of the FeH system. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Computational Materials Science & Engineering, Modeling and Simulation, Machine Learning |