Algorithm 1 ICBPC algorithm

Input:X = unlabeled images;

Output:K classes of clustering images;

1: Initialize PEDCC cluster centers;

2: repeat

3:    $\widehat{X}$ = Augumentation(X);

4:    $\widehat{Z}$ = Encoder($\widehat{X}$); Z = Encoder(X);

5:    $loss1$ = MMD($\mathrm{Z}\odot \widehat{\mathrm{Z}}$, PEDCC); $loss2$ = Contrastive loss($\mathrm{Z}$, $\widehat{\mathrm{Z}}$);

6: until Stopping criterion meet