PCLR Fold 4

Back to PCLR overview · Back to main index

Fold metrics

Metric	Value
Test accuracy	5.54%
Test F1 score	0.0604
Hierarchical loss	0.74792094
P-adic loss (total)	3952.37949101
P-adic loss (mean)	0.85235702
Prime base	83
Training samples	18,460
Test samples	4,637
Trained at	2026-10-13T05:40:46+11:00

P-adic loss breakdown

Agreement	Count	Share	Cost per mistake	Total contribution
Exact match	257	5.54%	0.000000	0.000000
p^4	2	0.04%	0.000000	0.000000
p^3	26	0.56%	0.000002	0.000045
p^2	290	6.25%	0.000145	0.042096
p^1	111	2.39%	0.012048	1.337349
p^0	3,951	85.21%	1.000000	3951.000000

Tag Rank vs Coefficient Magnitude

Tag rank vs max coefficient magnitude — Scatter plot showing the relationship between tag battle ranking and maximum absolute coefficient value across all taxonomies. Shows which tags have the strongest influence in the model.

About p-adic loss

P-adic loss measures the distance between predicted and true taxonomy using p-adic metric (base 83). Lower values indicate closer predictions in the taxonomy hierarchy. This metric is shared with the umllr model for comparison.