κ-means for streaming and distributed big sparse data

We provide the first streaming algorithm for computing a provable approximation to the κ-means of sparse Big Data. Here, sparse Big Data is a stream of n vectors in ℝd, where each vector has O(1) non-zeroes entries and possibly d ≥ n. E.g., adjacency matrix of a graph, web-links, social network, document-terms, or image-features matrices. Our streaming algorithm stores at most logn κ^O(1) input points in memory. If the stream is distributed among M machines, the running time reduces by a factor of M, while communicating a total of M κ^O(1) (sparse) input points between the machines. Our main contribution is a deterministic algorithm for computing a sparse (κ,ϵ)-coreset, which is a weighted subset of κ^O(1) input points that approximates the sum of squared distances from the n input points to every set of κ centers, up to (1 ± ϵ) factor, for any given constant ϵ > 0. This is the first such coreset of size independent of both d and n. Our experimental results show how our algorithm can bs used to boost the performance of any given κ-means heuristics, even in the off-line setting. Open access to our implementation is also provided.