This repository contains solution methods for the inverse heat problem: From sparse temperature observation of a ground truth forward simulation, the problem's parameters, such as initial temperature distribution, diffusivity, or heat source strength are inferred. Currently, there are two methods implemented using JAX: MCMC sampling and differentiable simulations.
This is just a toy repository where I test out some solution methods for inverse problems in general. In the future, I also want to implement:
- PINNs
- Neural Operators
- SG-MCMC