New bounds for k-means and information k-means
share
In this paper, we derive a new dimension-free non-asymptotic upper bound for the quadratic k-means excess risk related to the quantization of an i.i.d sample in a separable Hilbert space. We improve the bound of order 𝒪( k / √(n)) of Biau, Devroye and Lugosi, recovering the rate √(k/n) that has already been proved by Fefferman, Mitter, and Narayanan and by Klochkov, Kroshnin and Zhivotovskiy but with worse log factors and constants. More precisely, we bound the mean excess risk of an empirical minimizer by the explicit upper bound 16 B^2 log(n/k) √(k log(k) / n), in the bounded case when ℙ( ‖ X ‖≤ B) = 1. This is essentially optimal up to logarithmic factors since a lower bound of order 𝒪( √(k^1 - 4/d/n)) is known in dimension d. Our technique of proof is based on the linearization of the k-means criterion through a kernel trick and on PAC-Bayesian inequalities. To get a 1 / √(n) speed, we introduce a new PAC-Bayesian chaining method replacing the concept of δ-net with the perturbation of the parameter by an infinite dimensional Gaussian process. In the meantime, we embed the usual k-means criterion into a broader family built upon the Kullback divergence and its underlying properties. This results in a new algorithm that we named information k-means, well suited to the clustering of bags of words. Based on considerations from information theory, we also introduce a new bounded k-means criterion that uses a scale parameter but satisfies a generalization bound that does not require any boundedness or even integrability conditions on the sample. We describe the counterpart of Lloyd's algorithm and prove generalization bounds for these new k-means criteria.
READ FULL TEXT