Back to ULR overview · Back to main index
| Metric | Value |
|---|---|
| Test accuracy | 53.63% |
| Test F1 score | 0.5735 |
| Hierarchical loss | 0.95784884 |
| P-adic loss (total) | 142.53948363 |
| P-adic loss (mean) | 0.20717948 |
| Prime base | 71 |
| Number of tags (input features) | 1,640 |
| Non-zero parameters | 1,685 / 359,379 (99.5% sparse) |
| L1 regularization (C) | 1.0000 |
| Training samples | 2,735 |
| Test samples | 688 |
| Agreement | Count | Share | Cost per mistake | Total contribution |
|---|---|---|---|---|
| Exact match | 369 | 53.63% | 0.000000 | 0.000000 |
| p^6 | 1 | 0.15% | 0.000000 | 0.000000 |
| p^5 | 0 | 0.00% | 0.000000 | 0.000000 |
| p^4 | 9 | 1.31% | 0.000000 | 0.000000 |
| p^3 | 38 | 5.52% | 0.000003 | 0.000106 |
| p^2 | 92 | 13.37% | 0.000198 | 0.018250 |
| p^1 | 37 | 5.38% | 0.014085 | 0.521127 |
| p^0 | 142 | 20.64% | 1.000000 | 142.000000 |
L1 (Lasso) regularization promotes sparsity by driving many coefficients to exactly zero. This model uses ALL available tags (1,640) but L1 regularization selects which features are actually used. The number of non-zero parameters (1,685) indicates how many coefficients the model actually uses.