[mlpack-git] (blog) master: Add week 3. (ef8b5e3)

gitdub at mlpack.org gitdub at mlpack.org
Mon Jun 13 12:43:37 EDT 2016


Repository : https://github.com/mlpack/blog
On branch  : master
Link       : https://github.com/mlpack/blog/compare/f2dfdbb794dc318ba5de69acae94c6fb6d6a52eb...ef8b5e3c13e981f189ed5304ab2a751738c1ba94

>---------------------------------------------------------------

commit ef8b5e3c13e981f189ed5304ab2a751738c1ba94
Author: MarcosPividori <marcos.pividori at gmail.com>
Date:   Mon Jun 13 13:43:37 2016 -0300

    Add week 3.


>---------------------------------------------------------------

ef8b5e3c13e981f189ed5304ab2a751738c1ba94
 content/blog/MarcosWeekThree.md | 35 +++++++++++++++++++++++++++++++++++
 1 file changed, 35 insertions(+)

diff --git a/content/blog/MarcosWeekThree.md b/content/blog/MarcosWeekThree.md
new file mode 100644
index 0000000..ecd5ac0
--- /dev/null
+++ b/content/blog/MarcosWeekThree.md
@@ -0,0 +1,35 @@
+tle: Approximate Nearest Neighbor Search - Week 3
+Date: 2016-06-13 21:00:00
+Tags: gsoc, knn, kfn
+Author: Marcos Pividori
+
+Last week, I finished the extension for approximate neighbor search
+[[1]](http://github.com/mlpack/mlpack/pull/684).
+
+Mainly, I modified the code to include an epsilon value, which represents the
+accepted relative error. It is considered by the prune rules when deciding if
+a given node combination should be analyzed.
+
+When doing dual tree search, the best between the modified $B_1$ bound (with
+epsilon) and the original $B_2$ bound is chosen. As was discussed in
+[[2]](https://github.com/mlpack/mlpack/pull/684#discussion_r66614123),
+with actual implementation, $B_2$ can not be relaxed to include the epsilon
+value.
+
+Then, I added many test cases with different values of epsilon and different
+kind of trees (KDTree/BallTree/CoverTree), checking the relative error between
+the approximated results and the true best candidates.
+
+The command line tools: mlpack_knn and lmpack_kfn, were updated to include an
+extra option "-e", to specify the relative error (default value: 0).
+We have been discussing about which approximation parameters use for KFN in
+[[3]](http://github.com/mlpack/mlpack/pull/684/files/07879a2cc79b35b10d7fae687d6e27ad90a9f2d7#r66611928).
+
+
+Also, I have been working improving existing code for NSModel, as was suggested
+by Sumedh in [[4]](http://github.com/mlpack/mlpack/issues/674). After
+considering many options, boost variant seems to be the most appropiate.
+
+Next week, I plan to continue working improving existing code and checking some
+details. Also, I will do some comparisons in performance of exact vs approximate
+neighbor search.




More information about the mlpack-git mailing list