Back to ULR overview · Back to main index
| Metric | Value |
|---|---|
| Test accuracy | 43.56% |
| Test F1 score | 0.4813 |
| Hierarchical loss | 0.89061311 |
| P-adic loss (total) | 1284.06232743 |
| P-adic loss (mean) | 0.28177800 |
| Prime base | 83 |
| Number of tags (input features) | 7,726 |
| Non-zero parameters | 10,517 / 7,657,457 (99.9% sparse) |
| L1 regularization (C) | 1.0000 |
| Training samples | 18,540 |
| Test samples | 4,557 |
| Agreement | Count | Share | Cost per mistake | Total contribution |
|---|---|---|---|---|
| Exact match | 1,985 | 43.56% | 0.000000 | 0.000000 |
| p^6 | 1 | 0.02% | 0.000000 | 0.000000 |
| p^5 | 6 | 0.13% | 0.000000 | 0.000000 |
| p^4 | 84 | 1.84% | 0.000000 | 0.000002 |
| p^3 | 279 | 6.12% | 0.000002 | 0.000488 |
| p^2 | 592 | 12.99% | 0.000145 | 0.085934 |
| p^1 | 330 | 7.24% | 0.012048 | 3.975904 |
| p^0 | 1,280 | 28.09% | 1.000000 | 1280.000000 |
L1 (Lasso) regularization promotes sparsity by driving many coefficients to exactly zero. This model uses ALL available tags (7,726) but L1 regularization selects which features are actually used. The number of non-zero parameters (10,517) indicates how many coefficients the model actually uses.