[PEPr] Comment on RFC::VectorIndex

Jesus M. Castagnetto (http://pear.php.net/user/jmcastagnetto) has commented on the proposal for RFC::VectorIndex. Comment: I agree that it should not be an RFC, but a Draft or a Proposal. As for the vector handling. The base class for Math_Vector is n-dimensional, the specialized cases of 2- and 3-D are there mostly for ease of handling. The undelying structure is a tuple, so it poses no limits on the size of the vector. The same goes for most vector operations, except that there is a need for a generalization of the dot product and cross product methods. BTW, you may want to take a look at the limit cases to avoid unstable solutions due to round-off and the like. I know that I ought to write some decent end-user docs for Math_Vector, as well as do a stable release (now that people had been using it with minimal number of bugs for over 2 years :-) I saw back in 2003 a talk by the author of the article you refer on Latent Semantic Indexing applied to PubMed (it was a Bioinformatics conference), and I was very impressed by his results. One thing you ought to be aware is that the generation of the vector representations and other stuff related to the method, is time consuming, searching shouldn't be. Proposal information: http://pear.php.net/pepr/pepr-proposal-show.php?id=414 -- Sent by PEPr, the automatic proposal system at http://pear.php.net