Merge components more efficiently

miplearn/branching.py

@@ -50,9 +50,10 @@ class BranchPriorityComponent(Component):
         pass
 
     
-    def merge(self, other):
-        if other.priority is not None:
-            self._merge(other.priority)
+    def merge(self, other_components):
+        for comp in other_components:
+            if comp.priority is not None:
+                self._merge(comp.priority)
             
             
     def _merge(self, priority):

miplearn/problems/tests/test_knapsack.py

@@ -66,7 +66,7 @@ def test_knapsack_fixed_weights_jitter():
                                  fix_w=True,
                                  w_jitter=randint(low=0, high=1001),
                                 )
-    instances = gen.generate(1_000)
+    instances = gen.generate(5_000)
     w = [instance.weights[0,0] for instance in instances]
     assert round(np.std(w), -1) == 290.
     assert round(np.mean(w), -2) == 1500.

miplearn/solvers.py

@@ -94,12 +94,14 @@ class LearningSolver:
             for instance in tqdm(instances, desc=label, ncols=80)
         )
         
-        solvers = [p[0] for p in solver_result_pairs]
+        subsolvers = [p[0] for p in solver_result_pairs]
         results = [p[1] for p in solver_result_pairs]
         
         for (name, component) in self.components.items():
-            for subsolver in solvers:
-                self.components[name].merge(subsolver.components[name])
+            subcomponents = [subsolver.components[name]
+                             for subsolver in subsolvers
+                             if name in subsolver.components.keys()]
+            self.components[name].merge(subcomponents)
         
         return results
 
@@ -122,4 +124,4 @@ class LearningSolver:
                 if component_name not in self.components.keys():
                     continue
                 else:
-                    self.components[component_name].merge(component)
+                    self.components[component_name].merge([component])

miplearn/warmstart.py

@@ -180,13 +180,20 @@ class WarmStartComponent(Component):
             self.predictors[category] = deepcopy(self.predictor_prototype)
             self.predictors[category].fit(x_train, y_train)
 
-    def merge(self, other):
-        for c in other.x_train.keys():
-            if c not in self.x_train:
-                self.x_train[c] = other.x_train[c]
-                self.y_train[c] = other.y_train[c]
-            else:
-                self.x_train[c] = np.vstack([self.x_train[c], other.x_train[c]])
-                self.y_train[c] = np.vstack([self.y_train[c], other.y_train[c]])
-            if (c in other.predictors.keys()) and (c not in self.predictors.keys()):
-                self.predictors[c] = other.predictors[c]
+    def merge(self, other_components):
+        keys = set(self.x_train.keys())
+        for comp in other_components:
+            keys = keys.union(set(comp.x_train.keys()))
+            
+        for key in keys:
+            x_train_submatrices = [comp.x_train[key]
+                                   for comp in other_components
+                                   if key in comp.x_train.keys()]
+            y_train_submatrices = [comp.y_train[key]
+                                   for comp in other_components
+                                   if key in comp.y_train.keys()]
+            if key in self.x_train.keys():
+                x_train_submatrices += [self.x_train[key]]
+                y_train_submatrices += [self.y_train[key]]
+            self.x_train[key] = np.vstack(x_train_submatrices)
+            self.y_train[key] = np.vstack(y_train_submatrices)