Recent theoretical work has identified random projection as a promising dimensionality reduc-tion technique for learning mixtures of Gaus-sians. Here we summarize these results and il-lustrate them by a wide variety of experiments on synthetic and real data.
展开▼
