search

reedessick / universality

an investigation into the application of universal, EOS-independent relations in Gravitational Wave astronomy
MIT License
3 stars 2 forks source link

change library from pkl for processes to hdf5 #20

Closed reedessick closed 5 years ago

reedessick commented 5 years ago

each process should be a separate data set in the file with appropriate attributes

things like gpr-resample should produce this and things like gpr-gpr should read this in.

reedessick commented 5 years ago

change how all process files are read/written (standardize this within utils.py).

Similarly, we should give gpr-resample the ability to read in a csv file with hyperparameters and associated weights, and have it write an hdf5 with the appropriate weighted datasets contained therein. When plotting, we should overlay the inferred process for all hyperparameters supplied along with the "observed data". We should set the alpha values to scale with the associated weights, or something like that. We may need to muck with the plotting functions a bit to make this easy.

gpr-gpr should then know how to read in these hdf5 files, extract datasets, create conditioned processes, and write new hdf5 files with weighted data sets and updated weights as appropriate. Again, we should plot all processes used to condition the overarching process with alphas scaled by the weights.