Solving and predicting physic processes using machine learning algorithms
Date:
- Design a novel model, OpenPINN, combining deep learning and an unsupervised learning algorithm, to solve the physical processes, such as two-dimensional heat transfer within irregular domain;
- Implement several physical models into the library of OpenPINN, and increase the accuracy (98&\%&) and applicability of predictive models through algorithmic or architectural improvements;
- Work with application domain and high performance computing experts to ensure the required model fidelity while optimizing model training and execution performance on HPC platforms.
