DaikiTanaka-UT / JointOptimization

Hello Together,

In the below code you used 10 because CIFAR-10 has 10 classes, is it?

Lines 136 to 137 in 7e7a30d

    
           # We introduce a prior probability distribution p, which is a distribution of classes among all training data. 
        
           self.p = np.ones(10, dtype=np.float32)/10.

So, similarly for Clothing1M it would be 14 since it has 14 classes and If I have a dataset with n classes then we have to use n. Is my understanding correct here or is there something else I am missing?

	# We introduce a prior probability distribution p, which is a distribution of classes among all training data.
	self.p = np.ones(10, dtype=np.float32)/10.

Probability Distribution `p` in Loss Function