Read more about Vector Quantization
quantize — main compression utilityQuantizeOptions — hyper-parametersdequantize — restore compressed tables
to the original uncompressed number. Used for comparisons, plots, and export.compare_tables —
mean L2 difference between two alternative compressions
Can be used to compare with the original outputs.The conceptual Mermaid flowchart for the algorithm:
flowchart TD
upsc[/"Upscaled tables [sw,J,krw,krg]"/]
do_param_fit{Do parametric fit?}
embed{Fit total mobility?}
red{Use PCA?}
latent[/Feature vectors/]
clean[/De-duplicated data/]
repr[/N representative cluster points/]
tabs[/N representative saturation tables/]
upsc --> do_param_fit
do_param_fit -- no --> embed
embed -- no --> red
embed -- yes --> red
red -- no --> latent
red -- yes --> latent
do_param_fit-- Parametric LET fit for relative permeabilities --> red
latent -- Remove near-identical entries --> clean
clean -- "Use as is" --> repr
clean -- "Quantisation (k-means)" --> repr
repr -- "Inverse transform to tabular format" --> tabs