"Bias of importance measures for multi-valued attributes and solutions"

[James Bowery]

It seems the technique for constructing decision trees describe in:

http://enpub.fulton.asu.edu/hdeng3/MultiICANN2011.pdf

should already be represented in natural language compression algorithms under another name.

If so, what algorithms are they?  If not, why not?

[Matt Mahoney]

I'm not aware of decision trees used for data compression. They would probably not be useful for text or images, but might be for databases.

Also, the paper tests only on artificial data. I'm not convinced they have a useful result.

 

-- Matt Mahoney, matma...@yahoo.com

---

From: James Bowery <jabo...@gmail.com>
To: Hutter Prize <Hutter...@googlegroups.com>
Sent: Sunday, July 31, 2011 1:52 PM
Subject: [Hutter Prize] "Bias of importance measures for multi-valued attributes and solutions"

--
You received this message because you are subscribed to the Google Groups "Hutter Prize" group.
To post to this group, send email to hutter...@googlegroups.com.
To unsubscribe from this group, send email to hutter-prize...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/hutter-prize?hl=en.

[James Bowery]

OK, I guess the thing that drew my attention was the use of their "pForest" for context mixing.