Repository for the paper "Bayesian Quadrature on Riemannian Data Manifolds" (ICML 2021). See https://proceedings.mlr.press/v139/frohlich21a/frohlich21a.pdf
This code includes the core of the LAND mixture model (geodesic methods + optimization + quadrature) and a Bayesian quadrature (BQ) implementation based on the ''bayesquad'' library (see bayesquad/LICENSE). A small demo is included in the file demo.py, which shows how to fit a LAND mixture using MC or BQ.
Requirements:
- python==3.7.7
- gpy==1.9.9
- pymanopt==0.2.5
- scikit-learn==0.23.1
- scipy==1.5.2
- dill==0.3.3
- multimethod==1.4