Optimizing Correlation Analysis of Financial Market data Streams Using Intel® Math Kernel library Share your comment!

Bookmark and Share


This article demonstrates the superb performance advantages of Intel MKL in the implementation of the online noise filtration algorithm on Intel Xeon processors. It also demonstrates a straightforward port to the Intel Xeon Phi coprocessor. Although the problem of interest in this article is correlation analysis of financial market data, the principles and statistical analysis techniques used can find applications in many other fields, such as data mining, machine learning, pattern recognition, and bioinformatics. A common problem in these applications is transforming data in a highly dimensional space to a space with a reduced number of dimensions, i.e., dimensionality reduction. Principal component analysis (PCA) and the closely related linear discriminant analysis (LDA) are the most frequently used methods of dimensionality reduction. These methods require computation of statistical estimates for the raw data, such as variance, covariance, and correlation. They also typically involve linear transformation of large datasets. Intel MKL offers highly optimized and extensively threaded summary statistics functions and linear algebra functions on both Intel Xeon architectures and Intel Xeon Phi coprocessors, and should be considered the math library of choice in powering these data- oriented and compute-intensive applications.

Read the complete posting here.

Posted on by Zhang Zhang, Technical Consulting Engineer, Andrey Nikolaev, Software Architect, and Victoriya Kardakova, Software Development Engineer, Intel®