Center Loss Guided Prototypical Networks for Unbalance Few-Shot Industrial Fault Diagnosis

Algorithm 1

The center loss algorithm.

	Input: Training data {}. Initialization parameters in convolution layers. In loss layer the parameters W and {}. Hyperparameter and learning rate. T ← 0.
	Output: The parameters.
(1)	while not converge do
(2)	t ← t + 1.
(3)	Compute loss by .
(4)	Compute back spread fault for each
(5)	Update W by
(6)	Update for every by .
(7)	Update by
(8)	end while