[mlpack-git] (blog) master: Lozhnikov, week 7. (4433ca4)

[gitdub at mlpack.org]

Repository : https://github.com/mlpack/blog
On branch  : master
Link       : https://github.com/mlpack/blog/compare/2e8554d16793593251b6080720fc3dbef2e2f935...4433ca4b02d3ad75acb72405475824fb879ffbbb

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

commit 4433ca4b02d3ad75acb72405475824fb879ffbbb
Author: Mikhail Lozhnikov <lozhnikovma at gmail.com>
Date:   Tue Jul 12 12:03:26 2016 +0300

    Lozhnikov, week 7.


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

4433ca4b02d3ad75acb72405475824fb879ffbbb
 content/blog/LozhnikovWeekSeven.md | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/content/blog/LozhnikovWeekSeven.md b/content/blog/LozhnikovWeekSeven.md
new file mode 100644
index 0000000..3d4acca
--- /dev/null
+++ b/content/blog/LozhnikovWeekSeven.md
@@ -0,0 +1,12 @@
+Title: Implementation of tree types : Week 7
+Date: 2016-07-12 12:00:00
+Tags: gsoc, space trees, Vantage point tree
+Author: Mikhail Lozhnikov
+
+Last week, I have been working on the implementation of `HollowBallBound`. This class represents the area included between two concentric ball bounds. The radius of the inner ball may be equal to 0.
+
+The `HollowBallBound` class is intended to replace the original piece-wise ball bound of the vantage point tree. This bound allows to calculate the minimum distance to a point much faster than the original bound. One of the disadvantages is the loss the crucial property of the binary space tree: the use of `HollowBallBound` may lead to overlapping children.
+
+Since the vantage point tree is slightly different from the binary space tree, namely each intermediate node contains a point, I had to move the vantage point tree to a separate class. Unfortunately, that led to the duplication of the code.
+
+Moreover, I added some tests for the vantage point tree.