Why do we have to regularize twice here?

[buaacarzp]

LOUPE/loupe.py

Lines 440 to 444 in f0adf52

	fv2 = tf.reshape(fv2,[-1,self.cluster_size*self.feature_size])
	fv2 = tf.nn.l2_normalize(fv2,1)
	fv2 = tf.reshape(fv2,[-1,self.cluster_size*self.feature_size])
	fv2 = tf.nn.l2_normalize(fv2,1)

[xmy0916]

here