Change labels y to be +1/-1 (cf. comment #92 (comment)).

Author: William de Vazelhes

metric_learn/constraints.py

@@ -107,6 +107,6 @@ def wrap_pairs(X, constraints):
   c = np.array(constraints[2])
   d = np.array(constraints[3])
   constraints = np.vstack((np.column_stack((a, b)), np.column_stack((c, d))))
-  y = np.vstack([np.ones((len(a), 1)), np.zeros((len(c), 1))])
+  y = np.vstack([np.ones((len(a), 1)), - np.ones((len(c), 1))])
   pairs = X[constraints]
   return pairs, y

metric_learn/itml.py

@@ -54,10 +54,9 @@ def __init__(self, gamma=1., max_iter=1000, convergence_threshold=1e-3,
   def _process_pairs(self, pairs, y, bounds):
     pairs, y = check_X_y(pairs, y, accept_sparse=False,
                                       ensure_2d=False, allow_nd=True)
-    y = y.astype(bool)
 
     # check to make sure that no two constrained vectors are identical
-    pos_pairs, neg_pairs = pairs[y], pairs[~y]
+    pos_pairs, neg_pairs = pairs[y == 1], pairs[y == -1]
     pos_no_ident = vector_norm(pos_pairs[:, 0, :] - pos_pairs[:, 1, :]) > 1e-9
     pos_pairs = pos_pairs[pos_no_ident]
     neg_no_ident = vector_norm(neg_pairs[:, 0, :] - neg_pairs[:, 1, :]) > 1e-9
@@ -76,8 +75,7 @@ def _process_pairs(self, pairs, y, bounds):
     else:
       self.A_ = check_array(self.A0)
     pairs = np.vstack([pos_pairs, neg_pairs])
-    y = np.hstack([np.ones(len(pos_pairs)), np.zeros(len(neg_pairs))])
-    y = y.astype(bool)
+    y = np.hstack([np.ones(len(pos_pairs)), - np.ones(len(neg_pairs))])
     return pairs, y
 
 
@@ -100,7 +98,7 @@ def fit(self, pairs, y, bounds=None):
     """
     pairs, y = self._process_pairs(pairs, y, bounds)
     gamma = self.gamma
-    pos_pairs, neg_pairs = pairs[y], pairs[~y]
+    pos_pairs, neg_pairs = pairs[y == 1], pairs[y == -1]
     num_pos = len(pos_pairs)
     num_neg = len(neg_pairs)
     _lambda = np.zeros(num_pos + num_neg)

metric_learn/mmc.py

@@ -83,10 +83,9 @@ def fit(self, pairs, y):
   def _process_pairs(self, pairs, y):
     pairs, y = check_X_y(pairs, y, accept_sparse=False,
                                       ensure_2d=False, allow_nd=True)
-    y = y.astype(bool)
 
     # check to make sure that no two constrained vectors are identical
-    pos_pairs, neg_pairs = pairs[y], pairs[~y]
+    pos_pairs, neg_pairs = pairs[y == 1], pairs[y == -1]
     pos_no_ident = vector_norm(pos_pairs[:, 0, :] - pos_pairs[:, 1, :]) > 1e-9
     pos_pairs = pos_pairs[pos_no_ident]
     neg_no_ident = vector_norm(neg_pairs[:, 0, :] - neg_pairs[:, 1, :]) > 1e-9
@@ -107,8 +106,7 @@ def _process_pairs(self, pairs, y):
       self.A_ = check_array(self.A0)
 
     pairs = np.vstack([pos_pairs, neg_pairs])
-    y = np.hstack([np.ones(len(pos_pairs)), np.zeros(len(neg_pairs))])
-    y = y.astype(bool)
+    y = np.hstack([np.ones(len(pos_pairs)), - np.ones(len(neg_pairs))])
     return pairs, y
 
   def _fit_full(self, pairs, y):
@@ -128,7 +126,7 @@ def _fit_full(self, pairs, y):
     eps = 0.01        # error-bound of iterative projection on C1 and C2
     A = self.A_
 
-    pos_pairs, neg_pairs = pairs[y], pairs[~y]
+    pos_pairs, neg_pairs = pairs[y == 1], pairs[y == -1]
 
     # Create weight vector from similar samples
     pos_diff = pos_pairs[:, 0, :] - pos_pairs[:, 1, :]
@@ -244,7 +242,7 @@ def _fit_diag(self, pairs, y):
         dissimilar pairs
     """
     num_dim = pairs.shape[2]
-    pos_pairs, neg_pairs = pairs[y], pairs[~y]
+    pos_pairs, neg_pairs = pairs[y == 1], pairs[y == -1]
     s_sum = np.sum((pos_pairs[:, 0, :] - pos_pairs[:, 1, :]) ** 2, axis=0)
 
     it = 0

metric_learn/sdml.py

@@ -135,5 +135,4 @@ def fit(self, X, y, random_state=np.random):
     pos_neg = c.positive_negative_pairs(num_constraints,
                                               random_state=random_state)
     pairs, y = wrap_pairs(X, pos_neg)
-    y = 2 * y - 1
     return SDML.fit(self, pairs, y)