Back to ULR overview · Back to main index
| Metric | Value |
|---|---|
| Test accuracy | 45.11% |
| Test F1 score | 0.4937 |
| Hierarchical loss | 0.89437208 |
| P-adic loss (total) | 761.73812951 |
| P-adic loss (mean) | 0.27892279 |
| Prime base | 79 |
| Number of tags (input features) | 4,981 |
| Non-zero parameters | 6,588 / 3,332,958 (99.8% sparse) |
| L1 regularization (C) | 1.0000 |
| Training samples | 10,828 |
| Test samples | 2,731 |
| Agreement | Count | Share | Cost per mistake | Total contribution |
|---|---|---|---|---|
| Exact match | 1,232 | 45.11% | 0.000000 | 0.000000 |
| p^6 | 2 | 0.07% | 0.000000 | 0.000000 |
| p^5 | 6 | 0.22% | 0.000000 | 0.000000 |
| p^4 | 50 | 1.83% | 0.000000 | 0.000001 |
| p^3 | 131 | 4.80% | 0.000002 | 0.000266 |
| p^2 | 339 | 12.41% | 0.000160 | 0.054318 |
| p^1 | 212 | 7.76% | 0.012658 | 2.683544 |
| p^0 | 759 | 27.79% | 1.000000 | 759.000000 |
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 (6,588) indicates how many coefficients the model actually uses.