Back to ULR overview · Back to main index
| Metric | Value |
|---|---|
| Test accuracy | 51.72% |
| Test F1 score | 0.5630 |
| Hierarchical loss | 0.90647627 |
| P-adic loss (total) | 419.44146083 |
| P-adic loss (mean) | 0.22957934 |
| Prime base | 79 |
| Number of tags (input features) | 3,664 |
| Non-zero parameters | 4,513 / 1,788,520 (99.7% sparse) |
| L1 regularization (C) | 1.0000 |
| Training samples | 7,283 |
| Test samples | 1,827 |
| Agreement | Count | Share | Cost per mistake | Total contribution |
|---|---|---|---|---|
| Exact match | 945 | 51.72% | 0.000000 | 0.000000 |
| p^6 | 2 | 0.11% | 0.000000 | 0.000000 |
| p^5 | 2 | 0.11% | 0.000000 | 0.000000 |
| p^4 | 32 | 1.75% | 0.000000 | 0.000001 |
| p^3 | 91 | 4.98% | 0.000002 | 0.000185 |
| p^2 | 226 | 12.37% | 0.000160 | 0.036212 |
| p^1 | 111 | 6.08% | 0.012658 | 1.405063 |
| p^0 | 418 | 22.88% | 1.000000 | 418.000000 |
L1 (Lasso) regularization promotes sparsity by driving many coefficients to exactly zero. This model uses ALL available tags (3,664) but L1 regularization selects which features are actually used. The number of non-zero parameters (4,513) indicates how many coefficients the model actually uses.