ChanServ changed the topic of #mlpack to: "mlpack: a fast, flexible machine learning library :: We don't always respond instantly, but we will respond; please be patient :: Logs at http://www.mlpack.org/irc/
ImQ009 has joined #mlpack
< HavshxjdnaggzGit> How about if I want knn with Laplacian-kernel ?
< HavshxjdnaggzGit> I wrote this but it makes error
< HavshxjdnaggzGit> NeighborSearch<neighbor::KNN, mlpack::kernel::LaplacianKernel, arma::mat, mlpack::tree::VPTree> model(data);
< HavshxjdnaggzGit> This works for me
< HavshxjdnaggzGit> NeighborSearch < NearestNeighborSort, EuclideanDistance, arma::mat, mlpack::tree::RPTree> KNN(data);
< HavshxjdnaggzGit> but if I want the metric to be laplaciankernel it doesn
< HavshxjdnaggzGit> doesn't work
< AakashkaushikGit> I had a question, it's not specifically regarding the implementation but as i saw every implementation of activation function carries an inverse of that function if anyone can explain why and where do we need the inverse of a activation function in a neural networks computation, that will be helpful.
< rcurtin> HavshxjdnaggzGit: the laplacian kernel is not a metric; if you are looking to find points witj the maximum kerbel value, try FastMKS instead
The_LoudSpeaker has quit [Quit: Leaving bye!]
The_LoudSpeaker has joined #mlpack
< shrit[m]> rcurtin: The logo is good, it is not noticeable the rotation in the photo. Maybe in reality it is different, but the overall is perfect. It will be interesting to have ensmallen logo also, but I do not know how
< rcurtin> shrit[m]: yeah, I will have to talk to Walker at NumFOCUS about also making ensmallen stuff available
< rcurtin> also, I got my 386 running potato :) it turns out it's *way* quicker to upgrade distributions when you haven't installed any packages on it yet :)
< rcurtin> now let's see if I can get it running woody :)
< abernauer[m]> Grats everyone on passing their GSOC goals for their projects.
ImQ009 has quit [Quit: Leaving]
< rcurtin> shrit[m]: just noticed that *all* of the cereal mlpack_test tests are passing! awesome work
< HavshxjdnaggzGit> I want to measure the similarity between points using laplacian. I know it is a metric for similarity.
< rcurtin> HavshxjdnaggzGit: yes, but it's not a distance metric, which is what KNN requires
< rcurtin> I think that FastMKS will do what you're looking to do
< rcurtin> now, depending on your data, you may find that just a brute-force computation of all similarities is quickest; you can set that by setting `naive = true` in the constructor to FastMKS
< rcurtin> but it's hard to say what will be quickest, so maybe try it out and see what happens :)
< rcurtin> shrit[m]: since those fixes to mlpack_catch_test look so simple, I'll go ahead and start reviewing the PR... it will take a bit to get through it all :-D
< shrit[m]> Yes of course