An Open Source C++ Implementation of Multi-Threaded Gaussian Mixture Models, k-Means and Expectation Maximisation

摘要

Modelling of multivariate densities is a core component in many signalprocessing, pattern recognition and machine learning applications. Themodelling is often done via Gaussian mixture models (GMMs), which usecomputationally expensive and potentially unstable training algorithms. Weprovide an overview of a fast and robust implementation of GMMs in the C++language, employing multi-threaded versions of the Expectation Maximisation(EM) and k-means training algorithms. Multi-threading is achieved throughreformulation of the EM and k-means algorithms into a MapReduce-like framework.Furthermore, the implementation uses several techniques to improve numericalstability and modelling accuracy. We demonstrate that the multi-threadedimplementation achieves a speedup of an order of magnitude on a recent 16 coremachine, and that it can achieve higher modelling accuracy than a previouslywell-established publically accessible implementation. The multi-threadedimplementation is included as a user-friendly class in recent releases of theopen source Armadillo C++ linear algebra library. The library is provided underthe permissive Apache~2.0 license, allowing unencumbered use in commercialproducts.