2012/5/6 David Warde-Farley <[email protected]>: > On 2012-05-05, at 6:28 PM, Vlad Niculae <[email protected]> wrote: > >> These need to explore as many of the possible gotchas as we can: wide X, >> tall X, sparse X, etc. Maybe use our generators. > > So, one use case I've used to benchmark some of my own code is Coates et al's > dataset of 400,000 CIFAR10 patches with 108 PCA-whitened dimensions (I'm > generally pretty interested in the regime involving dense features, hundreds > of thousands to millions of examples, and hundreds to low thousands of > features). > > I'd also suggest looking into the Cython-with-OpenMP situation on Windows to > determine whether the pragmas and whatnot will be cleanly ignored by the free > compilers (I.e. whether or not that leveraging Cython.parallel will hose the > build on Windows, like we discussed on a ticket).
Indeed: Vlad could you start a wiki page with a list of possible optimizations to explore (like is cython + openmp a maintainable approach that works both on posix and windows and which algorithms are the primary target for such //-isation strategy). -- Olivier http://twitter.com/ogrisel - http://github.com/ogrisel ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Scikit-learn-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
