Hello all,
I am working on a project that involves training a neural network to output covariance matrices. However, I am having trouble ensuring that the matrices are always positive semidefinite. I’ve read about techniques like Cholesky decomposition and matrix exponential, but I’m not sure how to implement them in PyTorch.
Has anyone here had experience with this issue and could provide some guidance on how to ensure that the network outputs positive semidefinite matrices? Any help or advice would be greatly appreciated.
Thank you.
Best regards,