SignatuRE

Code for AISTATS 2022 paper Amortised Likelihood-free Inference for Expensive Time-series Simulators with Signatured Ratio Estimation (SignatuRE)

Structure of project:

• analysis: contains notebook for analysing results
• data: contains pseudo-observations for OU, MA2, and GSE simulators
• inference: implements kernel classifiers
• models: simulators for OU, MA2, and GSE models
• utils: code for computing performance metrics, defining prior densities, defining embedding networks, and posterior sampling

Main script: traing_and_sample.py

I made use of the sbi package without modification for the neural ratio estimation method.

Paper

The paper corresponding to this code can be found on PMLR
Short videos summarising the idea of this work can be found on SlidesLive and on YouTube

Citation

Please use the following citation:

@InProceedings{pmlr-v151-dyer22a,
  title = 	 {{Amortised Likelihood-free Inference for Expensive Time-series Simulators with Signatured Ratio Estimation}},
  author =       {Dyer, Joel and Cannon, Patrick W. and Schmon, Sebastian M.},
  booktitle = 	 {Proceedings of The 25th International Conference on Artificial Intelligence and Statistics},
  pages = 	 {11131--11144},
  year = 	 {2022},
  editor = 	 {Camps-Valls, Gustau and Ruiz, Francisco J. R. and Valera, Isabel},
  volume = 	 {151},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {28--30 Mar},
  publisher =    {PMLR},
  pdf = 	 {https://proceedings.mlr.press/v151/dyer22a/dyer22a.pdf},
  url = 	 {https://proceedings.mlr.press/v151/dyer22a.html}
}