[mlpack-git] (blog) master: Lozhnikov, week 5. (f8f65d3)

gitdub at mlpack.org gitdub at mlpack.org
Mon Jun 27 11:44:41 EDT 2016


Repository : https://github.com/mlpack/blog
On branch  : master
Link       : https://github.com/mlpack/blog/compare/e599a31f0db49068397033b3e6c15e2e9f969ec5...f8f65d3ca1c67f8e1ac55ffb91528a9624d817da

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

commit f8f65d3ca1c67f8e1ac55ffb91528a9624d817da
Author: Mikhail Lozhnikov <lozhnikovma at gmail.com>
Date:   Mon Jun 27 18:44:41 2016 +0300

    Lozhnikov, week 5.


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

f8f65d3ca1c67f8e1ac55ffb91528a9624d817da
 content/blog/LozhnikovWeekFive.md | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/content/blog/LozhnikovWeekFive.md b/content/blog/LozhnikovWeekFive.md
new file mode 100644
index 0000000..33c1227
--- /dev/null
+++ b/content/blog/LozhnikovWeekFive.md
@@ -0,0 +1,8 @@
+Title: Implementation of tree types : Week 5
+Date: 2016-06-27 18:30:00
+Tags: gsoc, space trees, R+ tree, R++ tree, Vantage point tree
+Author: Mikhail Lozhnikov
+
+As I had planned in the previous post I did the refactoring of the R+/R++ tree. I added a template parameter SweepType to the RPlusTreeSplit class and implemented a sweep method that tries to partition an intermediate node with the minimum number of splits. All sweep techniques are designed in  tree-independent manner. In order to do that I introduced a template parameter SplitPolicy. This parameter helps to determine the subtree in which we should insert any particular child of an intermediate node that is being split. Also I fixed some errors.
+
+Moreover, I began the vantage point tree implementation. Right now the split method (VantagePointSplit class) is implemented. I have to implement a piecewise ball boundary type since the BallBound is not suitable for the vantage point tree. Also I did not think about vantage point tree-specific tests.




More information about the mlpack-git mailing list