search

jiwoncpark / h0rton

Deep Modeling of Strong Gravitational Time Delay Lenses for Bayesian Inference of the Hubble Constant
MIT License
10 stars 1 forks source link

Add BNN-relevant metrics #3

Closed jiwoncpark closed 4 years ago

jiwoncpark commented 5 years ago

Examples are per-point predictive probability (PPPP) and pred RMSE. We should log them on tensorboard. Inference-specific metrics (H0) will come later in a different issue. Also, we should log the images of particularly bad outlier examples as images. Can consider logging plots as images as well, like what I've done here for a BNN in a different application.

jiwoncpark commented 5 years ago

For plots, mappings (true parameter vs. predicted mu \pm predicted sigma) on the validation set