What is the criteria for Conditional FCM algorithm (program) to terminate when used for centre selection in RBFNN?
one of the most used terminate rule is the center becoming no big change.
The criteria is that the objective function reaches local minima under specific constraints. That is to say when difference between two consecutive iterations for either the membership function or cluster center is less than predefined threshold. The other option is to reach a fixed number of iterations.
You may have a look in the following papers and this Matlab code:
https://www.researchgate.net/publication/287520512_Segmentation_of_Brain_Tissues_from_Magnetic_Resonance_Images_Using_Adaptively_Regularized_Kernel-Based_Fuzzy_C_-Means_Clustering
https://www.researchgate.net/publication/303801714_Robust_kernelized_local_information_fuzzy_C-means_clustering_for_brain_magnetic_resonance_image_segmentation
http://www.mathworks.com/matlabcentral/fileexchange/54141-arkfcm-algorithm
Article Segmentation of Brain Tissues from Magnetic Resonance Images...
Article Robust kernelized local information fuzzy C-means clustering...
You should 0.000005 or any other small fraction is the termination criteria in Fuzzy C-Means algorithm when using any ANN techniques.
- Badges
- Science topic
- Similar topics
- Filing