IVector - Fix LDA rank

With this MR we make sure that the dimension of the LDA matrix is not higher than its rank.

This possibly will impact the Vox forge error rates reported in bob.bio.spear and we need to fix that.

ping @vkrivokuca

bob/bio/gmm/algorithm/IVector.py

@@ -28,7 +28,8 @@ class IVector (GMM):
       use_lda = False,
       use_wccn = False,
       use_plda = False,
-      lda_dim = 50,
+      lda_dim = None,
+      lda_strip_to_rank=True,
       plda_dim_F  = 50,
       plda_dim_G = 50,
       plda_training_iterations = 50,
@@ -55,6 +56,7 @@ class IVector (GMM):
         use_wccn = use_wccn,
         use_plda = use_plda,
         lda_dim = lda_dim,
+        lda_strip_to_rank = lda_strip_to_rank,
         plda_dim_F  = plda_dim_F,
         plda_dim_G = plda_dim_G,
         plda_training_iterations = plda_training_iterations,
@@ -76,7 +78,7 @@ class IVector (GMM):
     self.whitening_trainer = bob.learn.linear.WhiteningTrainer()
 
     self.lda_dim = lda_dim
-    self.lda_trainer = bob.learn.linear.FisherLDATrainer(strip_to_rank=False)
+    self.lda_trainer = bob.learn.linear.FisherLDATrainer(strip_to_rank=lda_strip_to_rank)
     self.wccn_trainer = bob.learn.linear.WCCNTrainer()
     self.plda_trainer = bob.learn.em.PLDATrainer()
     self.plda_dim_F  = plda_dim_F
@@ -112,9 +114,16 @@ class IVector (GMM):
   def train_lda(self, training_features):
     logger.info("  -> Training LDA projector")
     self.lda, __eig_vals = self.lda_trainer.train(training_features)
+
     # resize the machine if desired
-    if self.lda_dim:
+    # You can only clip if the rank is higher than LDA_DIM 
+    if self.lda_dim is not None:
+      if len(__eig_vals) < self.lda_dim:
+        logger.warning("  -> You are resizing the LDA matrix to a value above its rank"
+                       "(from {0} to {1}). Be aware that this may lead you to imprecise eigenvectors.".\
+                        format(len(__eig_vals), self.lda_dim))
       self.lda.resize(self.lda.shape[0], self.lda_dim)
+       
 
   def train_wccn(self, training_features):
     logger.info("  -> Training WCCN projector")