Yes, unfortunately that's the nature of SVM. I However, there have been implementations with on-line learning capabilities for SVM that are (promised to be) nearly as accurate as the batch learning variant. Maybe the partial_fit could be a useful task for the GitHub "issue" list. Here would be some implementations of online learning SVMs that could be used for inspiration:
http://leon.bottou.org/projects/lasvm http://www.isn.ucsd.edu/svm/incremental/ http://www.cpdiehl.org/code.html http://people.eng.unimelb.edu.au/shiltona/svm/ Best, Sebastian > On Dec 15, 2014, at 9:52 AM, Andy <t3k...@gmail.com> wrote: > > On 12/14/2014 11:03 PM, Joel Nothman wrote: >> If the estimator supports `partial_fit`, you can use that, repeatedly, >> instead of `fit`. >> >> See documentation: >> http://scikit-learn.org/stable/modules/scaling_strategies.html >> http://scikit-learn.org/stable/auto_examples/cluster/plot_dict_face_patches.html >> > If the classifier does not have "partial_fit", as is the case for the > SVC, then there is no way > apart from refitting on the whole data, sorry. > > ------------------------------------------------------------------------------ > Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server > from Actuate! Instantly Supercharge Your Business Reports and Dashboards > with Interactivity, Sharing, Native Excel Exports, App Integration & more > Get technology previously reserved for billion-dollar corporations, FREE > http://pubads.g.doubleclick.net/gampad/clk?id=164703151&iu=/4140/ostg.clktrk > _______________________________________________ > Scikit-learn-general mailing list > Scikit-learn-general@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/scikit-learn-general ------------------------------------------------------------------------------ Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server from Actuate! Instantly Supercharge Your Business Reports and Dashboards with Interactivity, Sharing, Native Excel Exports, App Integration & more Get technology previously reserved for billion-dollar corporations, FREE http://pubads.g.doubleclick.net/gampad/clk?id=164703151&iu=/4140/ostg.clktrk _______________________________________________ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
