My feeling is that the feature template should be represented as a separate data structure from the data. The data is easily amendable to the scipy array representation. Alternatively, we can specify \phi(x,y) as a matrix, but I am not sure that is the most elegant way to make people work.
I am welcome to thoughts from others on that. On Wed, May 2, 2012 at 4:10 AM, Mathieu Blondel <[email protected]> wrote: > > > On Wed, May 2, 2012 at 4:47 PM, Olivier Grisel <[email protected]> > wrote: >> >> >> I think the main issue with building such a wrapper for crfsuite is >> that the internal memory layout of crfsuite cannot be directly mapped >> to numpy array nor scipy sparse matrices hence the wrapper needs to do >> a memory copy of the training set. > > > How do you represent features used by CRF efficiently with Numpy arrays and > Sparse matrices? The reason I'm asking is because most CRF implementations > use "feature templates" to automatically generate features on the fly. So, I > think we may have the same problem with CRFs that we have with HMMs : the > data representation is fundamentally different from the rest of the scikit. > > Mathieu > > ------------------------------------------------------------------------------ > 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 > ------------------------------------------------------------------------------ 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
