Automated algorithm selection in black-box optimization typically relies on supervised models that map landscape features to algorithm performance labels. Such models are costly to train, benchmark-dependent, and often fail to generalize to unseen problem classes. We study an unsupervised alternative: multi-kernel clustering over heterogeneous landscape representations, in which problem instances are grouped without using any performance labels, and the resulting clusters are mapped post hoc to solver recommendations through a strictly separated three-stage evaluation protocol. Drawing on two decades of advances in multiple kernel learning, we adopt a multi-kernel $k$-means formulation that jointly learns cluster assignments and kernel weights over four complementary landscape views: ELA, DeepELA, DoE2Vec, and TransOptAS. On affine BBOB-derived selector tasks for Differential Evolution and Particle Swarm Optimization, multi-kernel clustering consistently outperforms naive feature concatenation and the strongest single-view baseline under the EC-oriented selector metrics considered here. The learned kernel weights concentrate on ELA and TransOptAS, while DeepELA and DoE2Vec are pruned to zero weight on the studied tasks, providing a task-specific interpretation of which representations are retained by the multi-kernel model for selector-oriented grouping.