Localized multiple kernel learning is a promising strategy for combining multiple features or kernels in terms of their discriminative power. However, learning specific combination kernel for each sample generally leads to expensive computation and less reliability caused by some dominant samples. In addition, traditional sparse constraint generally causes that one of the best kernels is selected, while some useful kernels may have not been used efficiently. In this paper, we proposed a

more »

... ed non-sparse localized multiple kernel learning algorithm to tackle the issues above. In our algorithm, the samples are divided into groups and then the kernel weights are optimized for each group. Because the sparse constraint may lose useful kernels, we use an lp-norm constraint on the kernels and obtain non-sparse results to avoid losing useful kernels. The advantage of each kernel is adopted in various local spaces. Since some datasets consist of multiple features, we propose a multiple kernel clustering method to make the clustering result more reliable. Experiments state that our method performs better than other excellent MKL algorithms.