luchungi/sigker-nsdes

Code accompanying the paper "Non-adversarial training of Neural SDEs with signature kernel scores".

sigker-nsdes

This library is the accompanying code for the paper "Non-adversarial training of Neural SDEs with signature kernel scores".

Navigating the library

Each of the experiments in the paper can be run from the three notebooks within the /notebooks sub-directory:

1. Unconditional training of NSDEs: see the notebook unconditional_nsde.ipynb.
2. Conditional training of NSDEs: see the notebook conditional_nsde.ipynb.
3. Unconditional training of NSPDEs: see the notebook unconditional_nspde.ipynb.

Accompanying each of the above is an associated validation notebook, which contains the code used to generate the figures and statistics seen in the paper.