Schabby's Blog
Reserve Orbital Defence Commander

I wrote my Masters Thesis in the area of professional string and text fiddling. While the bigger part of the programming part of my thesis was done in Java, I used Matlab for the actual computing stuff and plotting. Java and character strings go well together as we know. However, Matlab and character strings are different story. The main problem with Matlab and string is probably a conceptual one: Matlab was mainly designed for technical computing, including fast linear algebra operations and has therefore a focus on matrix/vector data structures. Strings, however, are difficult to squeeze in a matrix/vector concept which makes it difficult to add it later to an existing (and backwards compatible) technical computing library.

Long story short, check out jBLAS, the fast linear algebra library for java.


Eine Antwort

  1. bistaumanga says:

    i performed a little test on my system:
    Matrix multiplication repeated 10 times for 1024*1024, 2048*2048 and 4096*4096
    here is the result:
    octave: 1.57 secs, 11.42 secs and 83.76 secs
    python(numpy): 1.36 secs, 5.67 secs and 22.78 secs
    jBLAS: 3.55 secs, 23.52 secs and 172 secs

    Is there something wrong in my system or this is the fastest i get on JAVA with jBLAS???

Post Comment

Please notice: Comments are moderated by an Admin.