DropRedundant: Collect data from multiple runs

miplearn/components/steps/drop_redundant.py

@@ -50,11 +50,9 @@ class DropRedundantInequalitiesStep(Component):
         self.current_iteration = 0
 
         logger.info("Predicting redundant LP constraints...")
-        cids = solver.internal_solver.get_constraint_ids()
-        x, constraints = self.x(
-            [instance],
-            constraint_ids=cids,
-            return_constraints=True,
+        x, constraints = self._x_test(
+            instance,
+            constraint_ids=solver.internal_solver.get_constraint_ids(),
         )
         y = self.predict(x)
 
@@ -84,11 +82,16 @@ class DropRedundantInequalitiesStep(Component):
         stats,
         training_data,
     ):
-        instance.slacks = solver.internal_solver.get_inequality_slacks()
-        stats["DropRedundant: Kept"] = self.total_kept
-        stats["DropRedundant: Dropped"] = self.total_dropped
-        stats["DropRedundant: Restored"] = self.total_restored
-        stats["DropRedundant: Iterations"] = self.total_iterations
+        if "slacks" not in training_data.keys():
+            training_data["slacks"] = solver.internal_solver.get_inequality_slacks()
+        stats.update(
+            {
+                "DropRedundant: Kept": self.total_kept,
+                "DropRedundant: Dropped": self.total_dropped,
+                "DropRedundant: Restored": self.total_restored,
+                "DropRedundant: Iterations": self.total_iterations,
+            }
+        )
 
     def fit(self, training_instances):
         logger.debug("Extracting x and y...")
@@ -100,34 +103,46 @@ class DropRedundantInequalitiesStep(Component):
                 self.classifiers[category] = deepcopy(self.classifier_prototype)
             self.classifiers[category].fit(x[category], y[category])
 
-    def x(self, instances, constraint_ids=None, return_constraints=False):
+    def _x_test(self, instance, constraint_ids):
         x = {}
         constraints = {}
+        cids = constraint_ids
+        for cid in cids:
+            category = instance.get_constraint_category(cid)
+            if category is None:
+                continue
+            if category not in x:
+                x[category] = []
+                constraints[category] = []
+            x[category] += [instance.get_constraint_features(cid)]
+            constraints[category] += [cid]
+        for category in x.keys():
+            x[category] = np.array(x[category])
+        return x, constraints
+
+    def _x_train(self, instances):
+        x = {}
         for instance in tqdm(
             InstanceIterator(instances),
             desc="Extract (rlx:drop_ineq:x)",
             disable=len(instances) < 5,
         ):
-            if constraint_ids is not None:
-                cids = constraint_ids
-            else:
-                cids = instance.slacks.keys()
+            for training_data in instance.training_data:
+                cids = training_data["slacks"].keys()
                 for cid in cids:
                     category = instance.get_constraint_category(cid)
                     if category is None:
                         continue
                     if category not in x:
                         x[category] = []
-                    constraints[category] = []
                     x[category] += [instance.get_constraint_features(cid)]
-                constraints[category] += [cid]
         for category in x.keys():
             x[category] = np.array(x[category])
-        if return_constraints:
-            return x, constraints
-        else:
         return x
 
+    def x(self, instances):
+        return self._x_train(instances)
+
     def y(self, instances):
         y = {}
         for instance in tqdm(
@@ -135,7 +150,8 @@ class DropRedundantInequalitiesStep(Component):
             desc="Extract (rlx:drop_ineq:y)",
             disable=len(instances) < 5,
         ):
-            for (cid, slack) in instance.slacks.items():
+            for training_data in instance.training_data:
+                for (cid, slack) in training_data["slacks"].items():
                     category = instance.get_constraint_category(cid)
                     if category is None:
                         continue

miplearn/components/steps/tests/test_drop_redundant.py

@@ -5,9 +5,18 @@
 import numpy as np
 from unittest.mock import Mock, call
 
-from miplearn import LearningSolver, Instance, InternalSolver
+from miplearn import (
+    LearningSolver,
+    Instance,
+    InternalSolver,
+    GurobiSolver,
+)
 from miplearn.classifiers import Classifier
-from miplearn.components.relaxation import DropRedundantInequalitiesStep
+from miplearn.components.relaxation import (
+    DropRedundantInequalitiesStep,
+    RelaxIntegralityStep,
+)
+from miplearn.problems.knapsack import GurobiKnapsackInstance
 
 
 def _setup():
@@ -115,14 +124,14 @@ def test_drop_redundant():
     )
 
     # LearningSolver calls after_solve
-    component.after_solve(solver, instance, None, {}, {})
+    training_data = {}
+    component.after_solve(solver, instance, None, {}, training_data)
 
     # Should query slack for all inequalities
     internal.get_inequality_slacks.assert_called_once()
 
     # Should store constraint slacks in instance object
-    assert hasattr(instance, "slacks")
-    assert instance.slacks == {
+    assert training_data["slacks"] == {
         "c1": 0.5,
         "c2": 0.0,
         "c3": 0.0,
@@ -130,7 +139,7 @@ def test_drop_redundant():
     }
 
 
-def test_drop_redundant_with_check_dropped():
+def test_drop_redundant_with_check_feasibility():
     solver, internal, instance, classifiers = _setup()
 
     component = DropRedundantInequalitiesStep(
@@ -195,12 +204,16 @@ def test_x_y_fit_predict_evaluate():
     )
 
     # First mock instance
-    instances[0].slacks = {
+    instances[0].training_data = [
+        {
+            "slacks": {
                 "c1": 0.00,
                 "c2": 0.05,
                 "c3": 0.00,
                 "c4": 30.0,
             }
+        }
+    ]
     instances[0].get_constraint_category = Mock(
         side_effect=lambda cid: {
             "c1": None,
@@ -218,12 +231,16 @@ def test_x_y_fit_predict_evaluate():
     )
 
     # Second mock instance
-    instances[1].slacks = {
+    instances[1].training_data = [
+        {
+            "slacks": {
                 "c1": 0.00,
                 "c3": 0.30,
                 "c4": 0.00,
                 "c5": 0.00,
             }
+        }
+    ]
     instances[1].get_constraint_category = Mock(
         side_effect=lambda cid: {
             "c1": None,
@@ -283,3 +300,71 @@ def test_x_y_fit_predict_evaluate():
     assert ev["True negative"] == 1
     assert ev["False positive"] == 1
     assert ev["False negative"] == 0
+
+
+def test_x_multiple_solves():
+    instance = Mock(spec=Instance)
+    instance.training_data = [
+        {
+            "slacks": {
+                "c1": 0.00,
+                "c2": 0.05,
+                "c3": 0.00,
+                "c4": 30.0,
+            }
+        },
+        {
+            "slacks": {
+                "c1": 0.00,
+                "c2": 0.00,
+                "c3": 1.00,
+                "c4": 0.0,
+            }
+        },
+    ]
+    instance.get_constraint_category = Mock(
+        side_effect=lambda cid: {
+            "c1": None,
+            "c2": "type-a",
+            "c3": "type-a",
+            "c4": "type-b",
+        }[cid]
+    )
+    instance.get_constraint_features = Mock(
+        side_effect=lambda cid: {
+            "c2": np.array([1.0, 0.0]),
+            "c3": np.array([0.5, 0.5]),
+            "c4": np.array([1.0]),
+        }[cid]
+    )
+
+    expected_x = {
+        "type-a": np.array(
+            [
+                [1.0, 0.0],
+                [0.5, 0.5],
+                [1.0, 0.0],
+                [0.5, 0.5],
+            ]
+        ),
+        "type-b": np.array(
+            [
+                [1.0],
+                [1.0],
+            ]
+        ),
+    }
+
+    expected_y = {
+        "type-a": np.array([[1], [0], [0], [1]]),
+        "type-b": np.array([[1], [0]]),
+    }
+
+    # Should build X and Y matrices correctly
+    component = DropRedundantInequalitiesStep()
+    actual_x = component.x([instance])
+    actual_y = component.y([instance])
+    print(actual_x)
+    for category in ["type-a", "type-b"]:
+        np.testing.assert_array_equal(actual_x[category], expected_x[category])
+        np.testing.assert_array_equal(actual_y[category], expected_y[category])