2014-03-16 16:56 GMT+01:00 Maheshakya Wijewardena <[email protected]>: > Hi Olivier, > > According to the documentation of the FLANN, among the neighbor search > methods, there is a LSH based approach. It uses multi-probe LSH. But that > type of index can only be used for matching binary features using Hamming > distances. So the LSH-ANN used in that is not usable in general setting. > > According to > Radim(http://radimrehurek.com/2014/01/performance-shootout-of-nearest-neighbours-querying/) > ANNOYs performance is much better than FLANN.
No, it rather says the opposite (look at the precision-vs-speed plots). Here is the conclusion by Radim: """ FLANN is faster, but Annoy is saner. I like how it has a single parameter that controls accuracy that “just works”. It offers pleasant accuracy/speed tradeoffs, a responsive developer, is memory lean, concise, supports mmap for sharing an index between processes (FLANN can only share the input matrix, and not the index itself). """ I agree with Radim that the focus on simple usage (a nicer hyperparameter to tradeoff precision with query speed) is nice if the raw speed is comparable to the other method that has a less user-friendly parameterization. -- Olivier http://twitter.com/ogrisel - http://github.com/ogrisel ------------------------------------------------------------------------------ Learn Graph Databases - Download FREE O'Reilly Book "Graph Databases" is the definitive new guide to graph databases and their applications. Written by three acclaimed leaders in the field, this first edition is now available. Download your free book today! http://p.sf.net/sfu/13534_NeoTech _______________________________________________ Scikit-learn-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
