Back to ULR overview · Back to main index
| Metric | Value |
|---|---|
| Test accuracy | 46.15% |
| Test F1 score | 0.5077 |
| Hierarchical loss | 0.89750437 |
| P-adic loss (total) | 714.94020878 |
| P-adic loss (mean) | 0.26083189 |
| Prime base | 79 |
| Number of tags (input features) | 4,981 |
| Non-zero parameters | 6,677 / 3,342,922 (99.8% sparse) |
| L1 regularization (C) | 1.0000 |
| Training samples | 10,818 |
| Test samples | 2,741 |
| Agreement | Count | Share | Cost per mistake | Total contribution |
|---|---|---|---|---|
| Exact match | 1,265 | 46.15% | 0.000000 | 0.000000 |
| p^6 | 3 | 0.11% | 0.000000 | 0.000000 |
| p^5 | 5 | 0.18% | 0.000000 | 0.000000 |
| p^4 | 47 | 1.71% | 0.000000 | 0.000001 |
| p^3 | 145 | 5.29% | 0.000002 | 0.000294 |
| p^2 | 336 | 12.26% | 0.000160 | 0.053838 |
| p^1 | 228 | 8.32% | 0.012658 | 2.886076 |
| p^0 | 712 | 25.98% | 1.000000 | 712.000000 |
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 (6,677) indicates how many coefficients the model actually uses.