Refactor EBM module to support TFP distributions

google / qhbm-library

Quantum Hamiltonian-Based Models built on TensorFlow Quantum

https://qhbm-library.readthedocs.io/en/latest/

Apache License 2.0

40 stars 15 forks source link

Refactor EBM module to support TFP distributions #106

Closed zaqqwerty closed 2 years ago

zaqqwerty commented 2 years ago

In the QHBM Library design doc, @geoffreyroeder described a refactor of the EBM module. The refactor has two main goals: unbundle model specification and model inference; have our probability distributions inherit from tfp.Distribution to allow routines from tfp to be applied directly to QHBMs.

zaqqwerty commented 2 years ago

This includes allowing for discrete flows.

zaqqwerty commented 2 years ago

qhbmlib now has structural support for user-specified TFP distributions: they can be embedded in EnergyInference layers, as Categorical and Bernoulli were in #111