OneofthemainchallengesinfeaturelearningusingDeepConvolutionalNeuralNetworks(DCNNs)forlarge-scalefacerecognitionisthedesignofappropriatelossfunc-tionsthatenhancediscriminativepower.Centrelosspe-nalisesthedistancebetweenthedeepfeaturesandtheircor-respondingclasscentresintheEuclideanspa