Discriminative cosine embedding in machine learning

A method, in a deep neural network, for training the network to learn an increased discrimination of feature vectors comprising:

• inputting a batch of training samples to the deep neural network;
• receiving a batch of feature vectors generated by the deep neural network;
• and
• refining the deep neural network by backpropagating a loss representing differences between the batch of feature vectors and ground truth results into the deep neural network, the loss determined by:
• a softmax loss function;
• and
• a Copernican loss comprising:
• a planetary loss function that minimizes intra-class variation of the feature vectors by minimizing a cosine distance of the feature vectors to their corresponding planet centers;
• and
• a sun loss function that maximizes the inter-class variation of the feature vectors by maximizing a cosine distance of the feature vectors away from a mean of the batch of training samples;
• wherein the loss is given by:
• Lc=Lprimary+λLp+Ls
• where:
• λ is a loss weight;
• Lprimary is the softmax loss function;
• Lp is the planetary loss function;
• and
• Ls is the sun loss function.