Randomized Construction APIs and their documentation.
A change after our last discussion is I separated get_distributions to get_distributions_from_dataset and get_distributions_from_json, where the former calculates distributions from the raw dataset and the latter returns distributions saved in a json file. The motivation is that calculating distributions from the raw dataset usually takes ~20 seconds and distributions don't change per dataset. get_distributions_from_json is like the shortcut to get the distributions.
Similar to train_test.json, we have official reconstruction_dataset_distribution and reconstruction_dataset_train_only_distribution. Not sure if we should also post them or just leave for users to generate their own jsons.
Randomized Construction APIs and their documentation.
A change after our last discussion is I separated
get_distributions
toget_distributions_from_dataset
andget_distributions_from_json
, where the former calculates distributions from the raw dataset and the latter returns distributions saved in a json file. The motivation is that calculating distributions from the raw dataset usually takes ~20 seconds and distributions don't change per dataset.get_distributions_from_json
is like the shortcut to get the distributions.Similar to
train_test.json
, we have officialreconstruction_dataset_distribution
andreconstruction_dataset_train_only_distribution
. Not sure if we should also post them or just leave for users to generate their own jsons.