Back to ULR overview · Back to main index
| Metric | Value |
|---|---|
| Test accuracy | 51.61% |
| Test F1 score | 0.5639 |
| Hierarchical loss | 0.95600702 |
| P-adic loss (total) | 253.32734462 |
| P-adic loss (mean) | 0.20363934 |
| Prime base | 71 |
| Number of tags (input features) | 2,797 |
| Non-zero parameters | 3,065 / 1,026,866 (99.7% sparse) |
| L1 regularization (C) | 1.0000 |
| Training samples | 4,937 |
| Test samples | 1,244 |
| Agreement | Count | Share | Cost per mistake | Total contribution |
|---|---|---|---|---|
| Exact match | 642 | 51.61% | 0.000000 | 0.000000 |
| p^6 | 2 | 0.16% | 0.000000 | 0.000000 |
| p^5 | 0 | 0.00% | 0.000000 | 0.000000 |
| p^4 | 17 | 1.37% | 0.000000 | 0.000001 |
| p^3 | 81 | 6.51% | 0.000003 | 0.000226 |
| p^2 | 158 | 12.70% | 0.000198 | 0.031343 |
| p^1 | 92 | 7.40% | 0.014085 | 1.295775 |
| p^0 | 252 | 20.26% | 1.000000 | 252.000000 |
L1 (Lasso) regularization promotes sparsity by driving many coefficients to exactly zero. This model uses ALL available tags (2,797) but L1 regularization selects which features are actually used. The number of non-zero parameters (3,065) indicates how many coefficients the model actually uses.