/inverse-probelms-urops

Primary LanguageJupyter Notebook

Using Deep Generative Priors For Inverse ProblemsAnd Bayesian Optimisation

There is a wiki over here for more details.

The basic struture of the YAML files inside config folder

exp_params:
  ... # Basic info that is constantly used by multiple functions, basically a dump right now

operator_params:
  ... # Info defining the operator

estimator_params:
  ... # This is where all the interesting things for the project happens actually

base_model_params:
  ... # This references an existing YAML that describes the base model

operator_params

Functions that are currently supported

  • operator: CenterOcclude/RandomOcclude/CompressedSensing

estimator_params

Functions that are currently supported, these are the path to config files

  • potential: mse/discriminator_weighted
  • initalisation: random/posterior/map_posterior
  • estimator: langevin/hmc/map/mala

base_model_params

Functions that are currently supported, these are the path to config files

  • model_name: mnist/gan/dcgan or mnist/vae/vanilla