Skilgannon's kd-tree

From RoboWiki
Jump to: navigation, search

is Julian Kent's kd tree faster than Rednaxela's or is it the other way around

Tmservo (talk)06:45, 4 October 2013

They're pretty much tied. Mine might be a little faster on systems where the data in the Kd-Tree is frequently pushed out of cache.

Skilgannon (talk)11:28, 4 October 2013

is third generation Rednaxela faster than Julian Kent

Tmservo (talk)11:38, 4 October 2013

No, as I said above, they are mostly the same except for cache conditions.

Skilgannon (talk)13:07, 4 October 2013

what about second generation Rednaxela and first generation Rednaxela

Tmservo (talk)21:22, 4 October 2013

3rd gen is the fastest of Rednaxela's trees. There's also a k-NN algorithm benchmark if you want to run some tests.

Voidious (talk)00:32, 5 October 2013

Voidious can you make a version of knn.jar with Julian Kent's tree and give instructions on how to use knn.jar on ubuntu 13.04

Tmservo (talk)00:51, 5 October 2013

I've never used it, sorry.

Why not just pick Skilgannon's or Rednaxela's and go with it? The speed difference will be negligible. Unless you're locked into some insane batch job on a supercomputer upon choosing, I reckon you'll spend more time trying to evaluate them than you'll save by using one over the other.

Voidious (talk)01:03, 5 October 2013

what's the difference between drussgt 3.0.0's kd-tree and drussgt 3.0.1's kd-tree

Tmservo (talk)15:29, 5 October 2013

No difference between them.

Skilgannon (talk)17:55, 5 October 2013

so what is the difference between drussgt 3.0.0 and drussgt 3.0.1

Tmservo (talk)18:01, 5 October 2013

3.0.0 had a small bug in the precise prediction code. Nothing to do with the Kd-Tree.

Skilgannon (talk)19:47, 5 October 2013
 
 
 
 

Try using this knn.jar, it has both trees as an option. Although that one isn't using my latest tree, you should be able to figure out how to change which tree it is using.

Seriously, the different in performance between RedGen3 and Skil are pretty negligible. You'd be better off putting the time into figuring out how to reduce a dimension of the data you're storing, or something like that, which will actually provide a speedup.

Skilgannon (talk)07:16, 5 October 2013

how do i change which tree its using

Tmservo (talk)13:28, 5 October 2013

You'll need to extract the .jar file, and make a new wrapper class in net/robowiki/knn/implementations directory, similar to the other ones there. Then at the top of net/robowiki/knn/implementations/KNNBenchmark.java you'll need to add a reference to the tree you want to test, like the others are referenced.

Skilgannon (talk)18:15, 5 October 2013

how do i make a new knn.jar with the latest skilgannon kd-tree

Tmservo (talk)18:40, 5 October 2013

You don't need to. Just run the KNNBenchmark by java KNNBenchmark.class

Skilgannon (talk)20:17, 5 October 2013

it doesn't work

Tmservo (talk)21:06, 5 October 2013

on my ubuntu i only have icedtea and openjdk

Tmservo (talk)21:10, 5 October 2013

What are the errors? IcedTea and OpenJDK should be fine.

Skilgannon (talk)22:13, 5 October 2013
 

i need anybody to give me test results of the latest skilgannon kd-tree

Tmservo (talk)22:16, 5 October 2013

Could not find or load main class KNNBenchmark.class

Tmservo (talk)22:16, 5 October 2013

You need to navigate to the base directory where you extracted the files. Then try java net.robowiki.knn.KNNBenchmark

Skilgannon (talk)23:47, 5 October 2013
 

Just curious, what are you using the tree for?

Skilgannon (talk)23:49, 5 October 2013

how do i put the latest skilgannon tree in KNN.jar

Tmservo (talk)00:01, 6 October 2013

because when i put it in the KNN jar it doesn't work

Tmservo (talk)00:03, 6 October 2013

This should be a good place to start [1]

Skilgannon (talk)07:26, 6 October 2013

i mean when i put drussgt 3.1.0's tree in the base directory where i extracted the files it doesn't work

Tmservo (talk)16:23, 6 October 2013

There were some minor changes in the tree, so the adapter might need to be changed from new KDTree(... to new KDTree.Euclidean(..., but after that you're on your own. You haven't given any indication of what is wrong, why it isn't working or even exactly what you did.

Skilgannon (talk)12:35, 7 October 2013

how do i change the adapter

Tmservo (talk)23:27, 7 October 2013

I already explained that here and in the post above.

Skilgannon (talk)06:54, 8 October 2013
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Personal tools