naywhayare changed the topic of #mlpack to: http://www.mlpack.org/ -- We don't respond instantly... but we will respond. Give it a few minutes. Or hours. -- Channel logs: http://www.mlpack.org/irc/
sumedh__ has joined #mlpack
sumedh_ has quit [Ping timeout: 240 seconds]
sumedh__ has quit [Ping timeout: 244 seconds]
< jenkins-mlpack>
Starting build #1984 for job mlpack - svn checkin test (previous build: SUCCESS)
< jenkins-mlpack>
* Ryan Curtin: Add comments to private functions in header file.
< jenkins-mlpack>
* Ryan Curtin: Refactor to avoid holding variables in the class itself (reduces the size of SA a bit). Remove an unnecessary variable too. Minor changes, no public API changes.
< jenkins-mlpack>
* Ryan Curtin: Add comments to private functions in header file.
< jenkins-mlpack>
* Ryan Curtin: Refactor to avoid holding variables in the class itself (reduces the size of SA a bit). Remove an unnecessary variable too. Minor changes, no public API changes.
< jenkins-mlpack>
* sumedhghaisas: * faster implementation of SVDBatchWithMomentum
< jenkins-mlpack>
* tolerance termination policy is modified according to new policy
< jenkins-mlpack>
* test point selection in validation RMSE termination is shifted to constructor
< jenkins-mlpack>
* Ryan Curtin: Rename parameters, clean up documentation; very minor tweaks. No functionality changes.
< jenkins-mlpack>
* Ryan Curtin: Rearrange parameters -- maxIterations is probably the most likely to be changed, and also maxIterations tends to be the first parameter for other algorithms.
< jenkins-mlpack>
* Ryan Curtin: Remove documentation for MoveDistributionType.
< jenkins-mlpack>
* andrewmw94: work arround for CMake bug.
< jenkins-mlpack>
* Ryan Curtin: Remove now-unnecessary include.
< jenkins-mlpack>
* Ryan Curtin: Create separate target for moving mlpack headers. This should fix #322.
< jenkins-mlpack>
* Ryan Curtin: Move LaplaceDistribution to mlpack::distribution and expand upon it, but then
< jenkins-mlpack>
remove MoveDistributionType template parameter from SA class because really, the
< jenkins-mlpack>
MoveControl() function is written specifically to work with the Laplace
< jenkins-mlpack>
distribution. I was incorrect to think it could be templatized so easily. The
< jenkins-mlpack>
implementation of the Laplace distribution has been inlined, too, which may help
< jenkins-mlpack>
it be a little faster.
sumedhghaisas has joined #mlpack
govg has joined #mlpack
sumedh_ has joined #mlpack
sumedhghaisas has quit [Ping timeout: 244 seconds]
udit has joined #mlpack
udit has quit [Client Quit]
govg has quit [Ping timeout: 272 seconds]
oldbeardo has joined #mlpack
< oldbeardo>
naywhayare: I'm running into problems while building the source with quic_svd included
< oldbeardo>
I'm getting multiple definition errors for all the CosineTree functions, something like multiple definition of `mlpack::tree::CosineTree::CosineTree(arma::Mat<double> const&)
< oldbeardo>
this happens when the CMake is linking files for bin/mlpack_test
< sumedh_>
oldbeardo: okay I maybe able to fix that... can I see the file where all these functions are defined??
< oldbeardo>
sumedh_: it's present in mlpack/core/tree/cosine_tree/
< sumedh_>
current svn version??
< oldbeardo>
sumedh_: it's probably because of the global class declaration, but I'm not sure
< oldbeardo>
not aware of that, let me check
< sumedh_>
cause last build of jenkins was a success... so I doubt they are present there...
< sumedh_>
oldbeardo: if not then make a commit ... I will update and check what I can do...
< oldbeardo>
sumedh_: svn version is 1.6.17
< oldbeardo>
sumedh_: okay, I will remove the file names from CMakeLists though, it shouldn't affect the build
< sumedh_>
ohh sorry.. I meant ... are those files committed ??
< sumedh_>
oldbeardo: can I view them in trunk version??
< oldbeardo>
sumedh_: they aren't in the trunk right now, I will commit in a few minutes
< sumedh_>
oldbeardo: okay... I will look into it... don't remove the files from CMake or I won't see the error
< oldbeardo>
sumedh_: I'm just commenting out relevant lines, I will tell you which ones to uncomment to see the error
< sumedh_>
okay that works...
< naywhayare>
oldbeardo: CosineTree isn't a templated class; its implementation should be in cosine_tree.cpp, not cosine_tree_impl.hpp
< oldbeardo>
naywhayare: crap, thanks :)
< naywhayare>
sorry that I didn't notice that earlier
< sumedh_>
naywhayare: inline-ing those will work right?? just asking...
< naywhayare>
sumedh_: yeah, that would work too
< naywhayare>
in this case it's probably better to just move it to a .cpp file
< sumedh_>
yes... obviously... I wanted to conform that my understanding is correct... :)
< oldbeardo>
naywhayare: you are up fairly early today!
< jenkins-mlpack>
Starting build #1985 for job mlpack - svn checkin test (previous build: SUCCESS)
govg has joined #mlpack
< naywhayare>
oldbeardo: actually, I woke up late :(
< oldbeardo>
naywhayare: well, the internet tells me it's around 9.30am over there, so I guess it's a matter of perspective :)
< naywhayare>
hah
< naywhayare>
I have been trying to wake up earlier for years, but I never have any luck...
< oldbeardo>
by the way, I committed the QUIC-SVD code, the build was successful
< naywhayare>
ok, thank you
< naywhayare>
looks like you accidentally added the files mlpack/trunk/src/mlpack/tests/CMakeLists.txt~ and mlpack/trunk/src/mlpack/methods/CMakeLists.txt~
< oldbeardo>
yeah, even I was thinking that addition of those files was strange, but somehow they were already there when I checked with 'svn status'
< oldbeardo>
I didn't add the files explicitly, how do I ensure that this doesn't happen again?
govg has quit [Quit: leaving]
< naywhayare>
what you probably did was 'svn add cosine_tree/' which added the directory and then all files in it recursively
< naywhayare>
you can use 'svn status' to take a look at which files are going to be committed, and then remove some of those with 'svn delete' if necessary
< oldbeardo>
okay, should I remove it now or have you already done it?
< naywhayare>
nah, you can go ahead and do it
< naywhayare>
I haven't done anything to those files
< naywhayare>
rann is 'rank-approximate nearest neighbor search'
< naywhayare>
so the solution isn't exact; it's approximate. and the algorithm decides whether or not to prune a branch based on how many points it has already sampled from that branch
< naywhayare>
so, Rescore() can be called and the algorithm can realize that it's already sampled enough points, and then it prunes the branch
< naywhayare>
and for rank-approximate nearest neighbor search pruning involves sampling more points, so queryIndex and referenceNode are both necessary
< andrewmw94>
ahh.
< andrewmw94>
I was like: "How can the score of a node change? It's just dependent on distance." I guess not.
< andrewmw94>
but it is for the BSP and R Trees
< andrewmw94>
right?
< andrewmw94>
(well, I know it is set to DBL_MAX when it is pruned, but other than that)
< andrewmw94>
because I assumed it was for my R Tree traversal code
< naywhayare>
the score depends entirely on the problem
< naywhayare>
so your traversal can't really assume anything about it (and doesn't need to)
< naywhayare>
the only thing is that the score is DBL_MAX when a prune should occur
< naywhayare>
and generally, a lower score means it's better to recurse into that node first
< andrewmw94>
well, I order the child nodes of a given node by their score to decide which to recurse into first. But if the scores can change that doesn't work.
< andrewmw94>
I'd have to re check it every time. Extra overhead, but if I have to I have to
< naywhayare>
in every implementation we have, the score can only change to DBL_MAX
< andrewmw94>
ok. Then I'll keep it as is.
< jenkins-mlpack>
Starting build #1990 for job mlpack - svn checkin test (previous build: SUCCESS)