jymesen-wang/2022-pr-ercan

Source Code for 'Entropy Regularization for Unsupervised Clustering with Adaptive Neighbors' (PR)

ERCAN

Using the code, please cite: Wang J, Ma Z, Nie F, et al. Entropy Regularization for Unsupervised Clustering with Adaptive Neighbors[J]. Pattern Recognition, volume 125, 2022, 108517, doi: 10.1016/j.patcog.2021.108517.

https://www.sciencedirect.com/science/article/pii/S0031320321006932

The code explanation: The main function of the code: ERCAN.m You can use demo.m to perform ERCAN clustering for Coil20 data set. If you have any questions, please connect [email protected]

Use of Main Function

[result,labelnew,r,t,ddd,S_sym] = ERCAN(X,label,k,gamma,eta,rate,iter)

example: [result,~] = ERCAN(X,label,3,1,10) Coil20 result ACC: 87.22% NMI: 94.50% Purity: 90.00% with k=3, eta=10.

Input: X: data matrix label: ground truth (for clustering result) k: the number of neighbors of similarity graph S \gamma: coefficient \gamma (if nargin=4, \gamma should be given by user) \eta: the multiplier of \gamma (if nargin>4, gamma is computed by \eta multiples ||dij_x||_{F}/n, and \gamma can be given any value) rate: the change rate of \Lambda, default 2 iter: Maximum of iteration, default 40