Talk:Pallas - Revision history

RednaxelaBot: Using .

2010-07-01T08:35:09Z

Using <syntaxhighlight>.

Nat at 08:32, 7 March 2009

2009-03-07T08:32:38Z

Rednaxela: List replies

2009-03-07T06:33:34Z

List replies

Nat at 06:17, 7 March 2009

2009-03-07T06:17:48Z

Nat at 09:43, 5 March 2009

2009-03-05T09:43:52Z

Rednaxela: KdTree explaination

2009-03-05T08:04:40Z

KdTree explaination

Nat at 07:53, 5 March 2009

2009-03-05T07:53:55Z

Rednaxela: DF-PIF, DF-GF, PM-PIF, PM-GF, etc

2009-03-05T07:49:27Z

DF-PIF, DF-GF, PM-PIF, PM-GF, etc

Nat at 07:28, 5 March 2009

2009-03-05T07:28:38Z

Nat at 07:24, 5 March 2009

2009-03-05T07:24:51Z

@@ Line 51: / Line 51: @@
 == ArrayList usage ==
 I'm wondering about highly used ArrayList. If I add one element to the array list each turn. will this faster?
-<pre>
+<syntaxhighlight>
 ArrayList list = new ArrayList(1000);
@@ Line 60: / Line 60: @@
      execute();
 } while(true);
-</pre>
+</syntaxhighlight>
 &raquo; <span style="font-size:0.9em;color:darkgreen;">[[User:Nat|Nat]] | [[User_talk:Nat|Talk]]</span> &raquo; 09:43, 5 March 2009 (UTC)

← Older revision		Revision as of 08:32, 7 March 2009
Line 69:		Line 69:

	Java's "Vector" is functionally like ArrayList except that it's "synchronized" which means it's safe for multi-threading, but at the cost of performance overhead. In other words.... don't use Vector. Also, if this is for storing a log like it's sounding like, I strongly suggest you avoid ArrayList, and use some kind of LinkedList (either a custom one or Java's one works). This is because for ArrayList to increase in size it must allocate new memory and copy the old contents over every so often, but a LinkedList has no such need. Therefore even though ArrayList tends to be faster on average usually, slowness spikes periodically are more likely to cause skipped turns. --[[User:Rednaxela\|Rednaxela]] 06:33, 7 March 2009 (UTC)		Java's "Vector" is functionally like ArrayList except that it's "synchronized" which means it's safe for multi-threading, but at the cost of performance overhead. In other words.... don't use Vector. Also, if this is for storing a log like it's sounding like, I strongly suggest you avoid ArrayList, and use some kind of LinkedList (either a custom one or Java's one works). This is because for ArrayList to increase in size it must allocate new memory and copy the old contents over every so often, but a LinkedList has no such need. Therefore even though ArrayList tends to be faster on average usually, slowness spikes periodically are more likely to cause skipped turns. --[[User:Rednaxela\|Rednaxela]] 06:33, 7 March 2009 (UTC)
		+
		+	I'm using custom LinkedList for almost all stats and highly use lists, but I just ask for some debugging graphics (DrawingBot), active waves and some not-so-large lists. One more, do generic-type list and raw-type list effect performance? I think not, but not sure. More, for active waves (including non-firing waves), which is better, ArrayList (as in BasicSurfer), LinkedList (as in Komarious) or PriorityQueue (sort by travalled distance)? I'm wonder about this bot speed because I requested many clusters to be created every tick =) » <span style="font-size:0.9em;color:darkgreen;">[[User:Nat\|Nat]] \| [[User_talk:Nat\|Talk]]</span> » 08:32, 7 March 2009 (UTC)

@@ Line 62: / Line 62: @@
 </pre>
 &raquo; <span style="font-size:0.9em;color:darkgreen;">[[User:Nat|Nat]] | [[User_talk:Nat|Talk]]</span> &raquo; 09:43, 5 March 2009 (UTC)
 == Performance issue ==
 I know that ABC use his own LinkedList for the performance. I'd want to know what is faster, ArrayList or Vactor. &raquo; <span style="font-size:0.9em;color:darkgreen;">[[User:Nat|Nat]] | [[User_talk:Nat|Talk]]</span> &raquo; 06:17, 7 March 2009 (UTC)

← Older revision		Revision as of 06:17, 7 March 2009
Line 62:		Line 62:
	</pre>		</pre>
	» <span style="font-size:0.9em;color:darkgreen;">[[User:Nat\|Nat]] \| [[User_talk:Nat\|Talk]]</span> » 09:43, 5 March 2009 (UTC)		» <span style="font-size:0.9em;color:darkgreen;">[[User:Nat\|Nat]] \| [[User_talk:Nat\|Talk]]</span> » 09:43, 5 March 2009 (UTC)
		+
		+	== Performance issue ==
		+	I know that ABC use his own LinkedList for the performance. I'd want to know what is faster, ArrayList or Vactor. » <span style="font-size:0.9em;color:darkgreen;">[[User:Nat\|Nat]] \| [[User_talk:Nat\|Talk]]</span> » 06:17, 7 March 2009 (UTC)

← Older revision		Revision as of 09:43, 5 March 2009
Line 48:		Line 48:

	Er, whoops, that was editing an old revision, not an edit conflict, thus I didn't get a warning (I clicked edit when looking at a diff). As as far as a KdTree, well, the 1st level could be the 1st dimension, the 2nd the 2nd and so on, BUT actually it's very often not done this way. The definition of a KdTree doesn't define which nodes split across which dimensions. In my impmentation I make nodes split at the dimension with the widest spread data, but there are plenty of other ways. And really, you split across dimensions more than once, there isn't really an actual limit on depth to a KdTree in theory. I suggest you look at some online resources on KdTrees, such as [http://www.cs.cmu.edu/~awm/animations/kdtree/ here]. The "nearest neighbor search" is for example is what a DC-gun does with a KdTree, and the "the levels of a kdtree" kind of shows the structure of the KdTree --[[User:Rednaxela\|Rednaxela]] 08:04, 5 March 2009 (UTC)		Er, whoops, that was editing an old revision, not an edit conflict, thus I didn't get a warning (I clicked edit when looking at a diff). As as far as a KdTree, well, the 1st level could be the 1st dimension, the 2nd the 2nd and so on, BUT actually it's very often not done this way. The definition of a KdTree doesn't define which nodes split across which dimensions. In my impmentation I make nodes split at the dimension with the widest spread data, but there are plenty of other ways. And really, you split across dimensions more than once, there isn't really an actual limit on depth to a KdTree in theory. I suggest you look at some online resources on KdTrees, such as [http://www.cs.cmu.edu/~awm/animations/kdtree/ here]. The "nearest neighbor search" is for example is what a DC-gun does with a KdTree, and the "the levels of a kdtree" kind of shows the structure of the KdTree --[[User:Rednaxela\|Rednaxela]] 08:04, 5 March 2009 (UTC)
		+
		+	== ArrayList usage ==
		+	I'm wondering about highly used ArrayList. If I add one element to the array list each turn. will this faster?
		+	<pre>
		+	ArrayList list = new ArrayList(1000);
		+
		+	do {
		+	// ....
		+	if ((list.size() - 200) < getTime())
		+	list.ensureCapacity(list.size() + 1000);
		+	execute();
		+	} while(true);
		+	</pre>
		+	» <span style="font-size:0.9em;color:darkgreen;">[[User:Nat\|Nat]] \| [[User_talk:Nat\|Talk]]</span> » 09:43, 5 March 2009 (UTC)

← Older revision		Revision as of 08:04, 5 March 2009
Line 46:		Line 46:

	Am I misunderstanding about KdTree? I think it a binary tree (of course). The 1th level is 1st dimension, 2nd is 2nd dimension and so. Is it mean that the actually depth of the tree is depth*dimension? I don't understand tree in anyway. » Nat \| Talk » 07:28, 5 March 2009 (UTC)		Am I misunderstanding about KdTree? I think it a binary tree (of course). The 1th level is 1st dimension, 2nd is 2nd dimension and so. Is it mean that the actually depth of the tree is depth*dimension? I don't understand tree in anyway. » Nat \| Talk » 07:28, 5 March 2009 (UTC)
		+
		+	Er, whoops, that was editing an old revision, not an edit conflict, thus I didn't get a warning (I clicked edit when looking at a diff). As as far as a KdTree, well, the 1st level could be the 1st dimension, the 2nd the 2nd and so on, BUT actually it's very often not done this way. The definition of a KdTree doesn't define which nodes split across which dimensions. In my impmentation I make nodes split at the dimension with the widest spread data, but there are plenty of other ways. And really, you split across dimensions more than once, there isn't really an actual limit on depth to a KdTree in theory. I suggest you look at some online resources on KdTrees, such as [http://www.cs.cmu.edu/~awm/animations/kdtree/ here]. The "nearest neighbor search" is for example is what a DC-gun does with a KdTree, and the "the levels of a kdtree" kind of shows the structure of the KdTree --[[User:Rednaxela\|Rednaxela]] 08:04, 5 March 2009 (UTC)

← Older revision		Revision as of 07:53, 5 March 2009
Line 42:		Line 42:

	That all seems fine to me. As to your formula, I'm too tired right now to really analyse it, but I would say it is depth, not depth/dimension. Each time you split it doesn't matter which dimension you split on, you still double the number of data point available. However, unless you get a perfect spread of data, some places will reach max depth before others, so you will never really 'fill' your tree. --[[User:Skilgannon\|Skilgannon]] 15:10, 3 March 2009 (UTC)		That all seems fine to me. As to your formula, I'm too tired right now to really analyse it, but I would say it is depth, not depth/dimension. Each time you split it doesn't matter which dimension you split on, you still double the number of data point available. However, unless you get a perfect spread of data, some places will reach max depth before others, so you will never really 'fill' your tree. --[[User:Skilgannon\|Skilgannon]] 15:10, 3 March 2009 (UTC)
		+
		+	I don't even think that it will have a really edit conflict on the new wiki. Rednaxela, I want to know that which editor do you use? I think the wiki editor warn you if there is edit conflict.
		+
		+	Am I misunderstanding about KdTree? I think it a binary tree (of course). The 1th level is 1st dimension, 2nd is 2nd dimension and so. Is it mean that the actually depth of the tree is depth*dimension? I don't understand tree in anyway. » Nat \| Talk » 07:28, 5 March 2009 (UTC)

@@ Line 14: / Line 14: @@
 And, if it is DC-PIF, what it is if it has only PIF? &raquo; <span style="font-size:0.9em;color:darkgreen;">[[User:Nat|Nat]] | [[User_talk:Nat|Talk]]</span> &raquo; 07:24, 5 March 2009 (UTC)
 == Simonton's KDTree ==
@@ Line 40: / Line 42: @@
 That all seems fine to me. As to your formula, I'm too tired right now to really analyse it, but I would say it is depth, not depth/dimension. Each time you split it doesn't matter which dimension you split on, you still double the number of data point available. However, unless you get a *perfect* spread of data, some places will reach max depth before others, so you will never really 'fill' your tree. --[[User:Skilgannon|Skilgannon]] 15:10, 3 March 2009 (UTC)

← Older revision		Revision as of 07:28, 5 March 2009
Line 40:		Line 40:

	That all seems fine to me. As to your formula, I'm too tired right now to really analyse it, but I would say it is depth, not depth/dimension. Each time you split it doesn't matter which dimension you split on, you still double the number of data point available. However, unless you get a perfect spread of data, some places will reach max depth before others, so you will never really 'fill' your tree. --[[User:Skilgannon\|Skilgannon]] 15:10, 3 March 2009 (UTC)		That all seems fine to me. As to your formula, I'm too tired right now to really analyse it, but I would say it is depth, not depth/dimension. Each time you split it doesn't matter which dimension you split on, you still double the number of data point available. However, unless you get a perfect spread of data, some places will reach max depth before others, so you will never really 'fill' your tree. --[[User:Skilgannon\|Skilgannon]] 15:10, 3 March 2009 (UTC)
		+
		+	Am I misunderstanding about KdTree? I think it a binary tree (of course). The 1th level is 1st dimension, 2nd is 2nd dimension and so. Is it mean that the actually depth of the tree is depth*dimension? I don't understand tree in anyway. » <span style="font-size:0.9em;color:darkgreen;">[[User:Nat\|Nat]] \| [[User_talk:Nat\|Talk]]</span> » 07:28, 5 March 2009 (UTC)

← Older revision		Revision as of 07:24, 5 March 2009
Line 12:		Line 12:

	I would call it Dynamic Clustering using Play-It-Forward. DC-PIF. DrussGT uses DC-GF. --[[User:Skilgannon\|Skilgannon]] 15:10, 3 March 2009 (UTC)		I would call it Dynamic Clustering using Play-It-Forward. DC-PIF. DrussGT uses DC-GF. --[[User:Skilgannon\|Skilgannon]] 15:10, 3 March 2009 (UTC)
		+
		+	And, if it is DC-PIF, what it is if it has only PIF? » <span style="font-size:0.9em;color:darkgreen;">[[User:Nat\|Nat]] \| [[User_talk:Nat\|Talk]]</span> » 07:24, 5 March 2009 (UTC)

	== Simonton's KDTree ==		== Simonton's KDTree ==