Back to ULR overview · Back to main index
| Metric | Value |
|---|---|
| Test accuracy | 46.46% |
| Test F1 score | 0.5079 |
| Hierarchical loss | 0.90347604 |
| P-adic loss (total) | 570.63035350 |
| P-adic loss (mean) | 0.20955944 |
| Prime base | 79 |
| Number of tags (input features) | 4,981 |
| Non-zero parameters | 6,725 / 3,357,868 (99.8% sparse) |
| L1 regularization (C) | 1.0000 |
| Training samples | 10,836 |
| Test samples | 2,723 |
| Agreement | Count | Share | Cost per mistake | Total contribution |
|---|---|---|---|---|
| Exact match | 1,265 | 46.46% | 0.000000 | 0.000000 |
| p^6 | 2 | 0.07% | 0.000000 | 0.000000 |
| p^5 | 2 | 0.07% | 0.000000 | 0.000000 |
| p^4 | 68 | 2.50% | 0.000000 | 0.000002 |
| p^3 | 160 | 5.88% | 0.000002 | 0.000325 |
| p^2 | 377 | 13.85% | 0.000160 | 0.060407 |
| p^1 | 282 | 10.36% | 0.012658 | 3.569620 |
| p^0 | 567 | 20.82% | 1.000000 | 567.000000 |
L1 (Lasso) regularization promotes sparsity by driving many coefficients to exactly zero. This model uses ALL available tags (4,981) but L1 regularization selects which features are actually used. The number of non-zero parameters (6,725) indicates how many coefficients the model actually uses.