scimba 1.3.1

This library implements some common tools for scientific machine learning

ScimBa

Scimba is a Python library that implements varying Scientific Machine Learning (SciML) methods for PDE problems, as well as tools for hybrid numerical methods.

The current version of the code solves parametric PDEs using various nonlinear approximation spaces such as neural networks, low-rank approximations, and nonlinear kernel methods. These methods:

• can handle complex geometries generated via level-set techniques and mappings, including sub-volumetric and surface domains;
• support function projections as well as elliptic, time-dependent, and kinetic parametric PDEs;
• are compatible with both space–time algorithms (PINN, Deep Ritz) and time-sequential ones (discrete PINNs, neural Galerkin and neural semi-Lagrangian schemes).

To achieve this, the code provides several optimization strategies, including:

• Adam, L-BFGS, SS-BFGS and SS-Broyden;
• natural gradient methods (for neural network-based models);
• hybrid least-squares approaches.

The current version of Scimba relies on a PyTorch backend.

A JAX version is under development and is released as an experimental module under scimba_jax. There is no warranty concerning the compatibility of the current release of scimba_jax with future versions. The given test coverage does not concern scimba_jax.

Documentation: https://www.scimba.org/

Code repository: https://gitlab.com/scimba/scimba/