Clustering is a central topic in unsupervised learning and its online formulation has received a lot of attention in recent years. In this paper, we study the classic facility location problem in the presence of multiple machine-learned advice. We design an algorithm with provable performance guarantees such that, if the advice is good, it outperforms the best-known online algorithms for the problem, and if it is bad it still matches their performance. We complement our theoretical analysis

more »

... an in-depth study of the performance of our algorithm, showing its effectiveness on synthetic and real-world data sets.