Back to ULR overview · Back to main index
| Metric | Value |
|---|---|
| Test accuracy | 53.69% |
| Test F1 score | 0.5716 |
| Hierarchical loss | 0.90730000 |
| P-adic loss (total) | 386.42538718 |
| P-adic loss (mean) | 0.24380151 |
| Prime base | 79 |
| Number of tags (input features) | 5,904 |
| Non-zero parameters | 4,561 / 3,230,035 (99.9% sparse) |
| L1 regularization (C) | 1.0000 |
| Training samples | 6,105 |
| Test samples | 1,585 |
| Agreement | Count | Share | Cost per mistake | Total contribution |
|---|---|---|---|---|
| Exact match | 854 | 53.88% | 0.000000 | 0.000000 |
| p^6 | 2 | 0.13% | 0.000000 | 0.000000 |
| p^5 | 4 | 0.25% | 0.000000 | 0.000000 |
| p^4 | 37 | 2.33% | 0.000000 | 0.000001 |
| p^3 | 66 | 4.16% | 0.000002 | 0.000134 |
| p^2 | 126 | 7.95% | 0.000160 | 0.020189 |
| p^1 | 111 | 7.00% | 0.012658 | 1.405063 |
| p^0 | 385 | 24.29% | 1.000000 | 385.000000 |
L1 (Lasso) regularization promotes sparsity by driving many coefficients to exactly zero. This model uses ALL available tags (5,904) but L1 regularization selects which features are actually used. The number of non-zero parameters (4,561) indicates how many coefficients the model actually uses.