Back to ULR overview · Back to main index
| Metric | Value |
|---|---|
| Test accuracy | 51.50% |
| Test F1 score | 0.5466 |
| Hierarchical loss | 0.90660013 |
| P-adic loss (total) | 371.08731066 |
| P-adic loss (mean) | 0.24656964 |
| Prime base | 79 |
| Number of tags (input features) | 5,904 |
| Non-zero parameters | 4,639 / 3,247,750 (99.9% sparse) |
| L1 regularization (C) | 1.0000 |
| Training samples | 6,185 |
| Test samples | 1,505 |
| Agreement | Count | Share | Cost per mistake | Total contribution |
|---|---|---|---|---|
| Exact match | 777 | 51.63% | 0.000000 | 0.000000 |
| p^6 | 2 | 0.13% | 0.000000 | 0.000000 |
| p^5 | 8 | 0.53% | 0.000000 | 0.000000 |
| p^4 | 44 | 2.92% | 0.000000 | 0.000001 |
| p^3 | 71 | 4.72% | 0.000002 | 0.000144 |
| p^2 | 149 | 9.90% | 0.000160 | 0.023874 |
| p^1 | 84 | 5.58% | 0.012658 | 1.063291 |
| p^0 | 370 | 24.58% | 1.000000 | 370.000000 |
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 (4,639) indicates how many coefficients the model actually uses.