http://ieeexplore.ieee.org TOC Alert for Publication# 5288520 2013 May 23 57 3/4 C1 C1 1380 57 3/4 C2 C2 6 57 3/4 1 2 52 57 3/4 0:1 0:4 60 57 3/4 1:1 1:13 2548 57 3/4 2:1 2:11 1932 57 3/4 3:1 3:10 877 57 3/4 4:1 4:11 950 57 3/4 5:1 5:10 1861 57 3/4 6:1 6:14 2257 57 3/4 7:1 7:11 1275 57 3/4 8:1 8:12 2416 $m$ -dimensional vector $y$) can be used to learn a dictionary matrix $A$ of dimension $m times k$, such that the documents can be approximately represented by a linear combination of a few columns of $A$. If a new document cannot be represented with low error as a sparse linear combination of these columns, then this is a strong indicator of novelty of the document. We scale up this approach to handle millions of documents by parallelizing sparse coding and dictionary learning, and by using the alternating-directions method to solve the resulting optimization problems. We conduct our experiments on high-performance computing clusters with differing architectures and evaluate our approach on news streams and streaming data from Twitter®. Based on the analysis, we share our insights on the distributed optimization and machine architecture that can help the design of exascale systems supporting data analytics.]]> 57 3/4 9:1 9:15 1476 57 3/4 10:1 10:6 99 57 3/4 11:1 11:13 6140 57 3/4 12:1 12:14 3986 57 3/4 13:1 13:12 10215 57 3/4 14:1 14:13 2289 57 3/4 15:1 15:10 9434 57 3/4 C3 C3 6