Scientific computing · Project 02
Physics-Informed Neural Networks
Exploring how physical laws can become part of a neural network's learning process when approximating solutions to differential equations.
- Field
- Scientific machine learning
- Status
- Research archive in progress
- Format
- Paper, experiments, and figures
- Role
- Details to be added
Introduction
Learning with physical constraints
Physics-informed neural networks connect approximation with prior structure. Instead of learning only from observed data, a model may also be guided by differential equations, boundary conditions, and other constraints that describe the system.
This introduction can later explain the exact equation, architecture, training process, evaluation, and where the method succeeded or failed. Technical evidence will remain in the document below.
Project document
Paper & experiment notes
The embedded PDF reserves the space for the final research report, technical appendix, or selected experimental results.