Torchquad

Torchquad is a Python module for numerical integration in arbitrary dimensions on the GPU, built on PyTorch. It was released by the Advanced Concepts Team at the European Space Agency, which says it is already used in internal projects there. Higher dimensional integration becomes very costly because the number of evaluation points grows exponentially with the number of dimensions, a problem known as the curse of dimensionality. It hits grid based methods hardest and also affects Monte Carlo methods, which need more points to converge. The team states that no public framework handled this problem on graphics processors, and torchquad was written to fill that gap, with CPU based tools as the comparison point in its evaluation notebook. It suits scientists and engineers who need to integrate high dimensional functions and want GPU speed without writing their own kernels. The module installs through the pip and conda package managers. The ESA page links to an example notebook, an evaluation notebook comparing it against standard CPU tools, GitHub pages and documentation. The page describes the project as released in version 0.1 and under constant development. The site publishes no pricing, and no paid plans are listed.