Merge pull request #284 from ntumlgroup/linear_multiclass

Add model name to linear models

Author: Eleven1Liu

libmultilabel/linear/linear.py

@@ -18,10 +18,12 @@
 
 
 class FlatModel:
-    def __init__(self, weights: np.matrix,
+    def __init__(self, name: str,
+                 weights: np.matrix,
                  bias: float,
                  thresholds: float | np.ndarray,
                  ):
+        self.name = name
         self.weights = weights
         self.bias = bias
         self.thresholds = thresholds
@@ -84,7 +86,8 @@ def train_1vsrest(y: sparse.csr_matrix,
         yi = y[:, i].toarray().reshape(-1)
         weights[:, i] = _do_train(2*yi - 1, x, options).ravel()
 
-    return FlatModel(weights=np.asmatrix(weights),
+    return FlatModel(name='1vsrest',
+                     weights=np.asmatrix(weights),
                      bias=bias,
                      thresholds=0)
 
@@ -169,15 +172,16 @@ def train_thresholding(y: sparse.csr_matrix,
         weights[:, i] = w.ravel()
         thresholds[i] = t
 
-    return FlatModel(weights=np.asmatrix(weights),
+    return FlatModel(name='thresholding',
+                     weights=np.asmatrix(weights),
                      bias=bias,
                      thresholds=thresholds)
 
 
 def _thresholding_one_label(y: np.ndarray,
-                           x: sparse.csr_matrix,
-                           options: str
-                           ) -> tuple[np.ndarray, float]:
+                            x: sparse.csr_matrix,
+                            options: str
+                            ) -> tuple[np.ndarray, float]:
     """Outer cross-validation for thresholding on a single label.
 
     Args:
@@ -223,10 +227,10 @@ def _thresholding_one_label(y: np.ndarray,
 
 
 def _scutfbr(y: np.ndarray,
-            x: sparse.csr_matrix,
-            fbr_list: list[float],
-            options: str
-            ) -> tuple[np.matrix, np.ndarray]:
+             x: sparse.csr_matrix,
+             fbr_list: list[float],
+             options: str
+             ) -> tuple[np.matrix, np.ndarray]:
     """Inner cross-validation for SCutfbr heuristic.
 
     Args:
@@ -414,15 +418,16 @@ def train_cost_sensitive(y: sparse.csr_matrix,
         w = _cost_sensitive_one_label(2*yi - 1, x, options)
         weights[:, i] = w.ravel()
 
-    return FlatModel(weights=np.asmatrix(weights),
+    return FlatModel(name='cost_sensitive',
+                     weights=np.asmatrix(weights),
                      bias=bias,
                      thresholds=0)
 
 
 def _cost_sensitive_one_label(y: np.ndarray,
-                             x: sparse.csr_matrix,
-                             options: str
-                             ) -> np.ndarray:
+                              x: sparse.csr_matrix,
+                              options: str
+                              ) -> np.ndarray:
     """Loop over parameter space for cost-sensitive on a single label.
 
     Args:
@@ -453,10 +458,10 @@ def _cost_sensitive_one_label(y: np.ndarray,
 
 
 def _cross_validate(y: np.ndarray,
-                   x: sparse.csr_matrix,
-                   options: str,
-                   perm: np.ndarray
-                   ) -> np.ndarray:
+                    x: sparse.csr_matrix,
+                    options: str,
+                    perm: np.ndarray
+                    ) -> np.ndarray:
     """Cross-validation for cost-sensitive.
 
     Args:
@@ -542,7 +547,8 @@ def train_cost_sensitive_micro(y: sparse.csr_matrix,
         w = _do_train(2*yi - 1, x, final_options)
         weights[:, i] = w.ravel()
 
-    return FlatModel(weights=np.asmatrix(weights),
+    return FlatModel(name='cost_sensitive_micro',
+                     weights=np.asmatrix(weights),
                      bias=bias,
                      thresholds=0)
 
@@ -590,7 +596,8 @@ def train_binary_and_multiclass(y: sparse.csr_matrix,
     # For labels not appeared in training, assign thresholds to -inf so they won't be predicted.
     thresholds = np.full(num_labels, -np.inf)
     thresholds[train_labels] = 0
-    return FlatModel(weights=np.asmatrix(weights),
+    return FlatModel(name='binary_and_multiclass',
+                     weights=np.asmatrix(weights),
                      bias=bias,
                      thresholds=thresholds)
 
@@ -615,7 +622,7 @@ def get_topk_labels(label_mapping: np.ndarray,
     """Get top k predictions from decision values.
 
     Args:
-        label_mapping (np.ndarray): A ndarray of class labels that maps each index (from 0 to ``num_class-1``) to its label. 
+        label_mapping (np.ndarray): A ndarray of class labels that maps each index (from 0 to ``num_class-1``) to its label.
         preds (np.ndarray): A matrix of decision values with dimension number of instances * number of classes.
         top_k (int): Determine how many classes per instance should be predicted.
 

libmultilabel/linear/tree.py

@@ -42,6 +42,7 @@ def __init__(self,
                  flat_model: linear.FlatModel,
                  weight_map: np.ndarray,
                  ):
+        self.name = 'tree'
         self.root = root
         self.flat_model = flat_model
         self.weight_map = weight_map

linear_trainer.py

@@ -5,8 +5,7 @@
 from tqdm import tqdm
 
 import libmultilabel.linear as linear
-from libmultilabel.common_utils import (argsort_top_k, dump_log,
-                                        is_multiclass_dataset)
+from libmultilabel.common_utils import argsort_top_k, dump_log
 from libmultilabel.linear.utils import LINEAR_TECHNIQUES
 
 
@@ -15,7 +14,7 @@ def linear_test(config, model, datasets):
         config.metric_threshold,
         config.monitor_metrics,
         datasets['test']['y'].shape[1],
-        multiclass=config.multiclass
+        multiclass=model.name=='binary_and_multiclass'
     )
     num_instance = datasets['test']['x'].shape[0]
 
@@ -73,7 +72,6 @@ def linear_run(config):
             config.label_file,
             config.include_test_labels,
             config.remove_no_label_data)
-        config.multiclass = is_multiclass_dataset(datasets['train'], label='y')
         model = linear_train(datasets, config)
         linear.save_pipeline(config.checkpoint_dir, preprocessor, model)