Fusion Multiple Kernel K-means

Multiple kernel clustering aims to seek an appropriate combination of base kernels mine inherent non-linear information for optimal clustering. Late fusion algorithms generate partitions independently and integrate them in the following procedure, improving overall efficiency. However, separate partition generation leads inadequate negotiation with procedure a great loss beneficial corresponding matrices, which negatively affects performance. To address this issue, we propose novel algorithm, termed as Fusion Kernel k-means (FMKKM), unifies learning late into one single objective function, adopts early technique capture more sufficient matrices. Specifically, helps keep details, further guides consensus stage, while provides positive feedback on two former procedures. The close collaboration three procedures results promising performance improvement. Subsequently, alternate optimization method convergence is developed solve resultant problem. Comprehensive experimental demonstrate that our proposed algorithm achieves state-of-the-art multiple public datasets, validating its effectiveness. code work publicly available at https://github.com/ethan-yizhang/Fusion-Multiple-Kernel-K-means.