home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.neural-nets
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!wupost!gumby!destroyer!ubc-cs!unixg.ubc.ca!kakwa.ucs.ualberta.ca!alberta!arms
- From: arms@cs.UAlberta.CA (Bill Armstrong)
- Subject: Re: Learning what COULD be learned
- Message-ID: <arms.711809827@spedden>
- Sender: news@cs.UAlberta.CA (News Administrator)
- Nntp-Posting-Host: spedden.cs.ualberta.ca
- Organization: University of Alberta, Edmonton, Canada
- References: <1992Jul7.074650.27125@aber.ac.uk> <13uievINN1mp@iraul1.ira.uka.de> <arms.711663417@spedden> <1992Jul21.082035.8898@aber.ac.uk> <arms.711759321@spedden> <1992Jul22.005147.2391@cs.wayne.edu>
- Date: Wed, 22 Jul 1992 12:57:07 GMT
- Lines: 34
-
- uds@wsu-eng.eng.wayne.edu (Seetamraju Udaybhaskar) writes:
-
- >In article <arms.711759321@spedden> arms@cs.UAlberta.CA (Bill Armstrong) writes:
- >>
- >>If a tree isn't learning the required task, as shown by a lack of
- >>further improvement, then you can double the size of the tree for that
- >>particular output and try again. Does this make it clear how the
- >>technique of independent trees would be used?
-
- >could you clarify, about how, with the number of inputs constant, the
- >size of the tree could be varied ?
-
- Sure. The use of the term tree is a misnomer if you also include the
- inputs, which must be in general connected to several nodes in the
- first layer of elements. Sorry for the terminological problem.
- To learn XOR, for example, you could start with at tree of two layers
- and synthesize x1* (~ x2) + x2 * (~x1). The input layer fans out to two
- nodes each. Those nodes are ANDs (*) and those ANDs output to an OR.
-
- What I have against non-trees is that there has to be a good reason
- for supposing that a node's learned output will serve two or more
- specific needs well. This seems to me to make learning much harder.
- But at the input level, there is no learning, so the argument doesn't apply.
- Besides, the only way to generate arbitrary functions is with repeated use of
- inputs (and complements in the case of ALNs).
-
- Hope that clears it up.
-
- Bill
- --
- ***************************************************
- Prof. William W. Armstrong, Computing Science Dept.
- University of Alberta; Edmonton, Alberta, Canada T6G 2H1
- arms@cs.ualberta.ca Tel(403)492 2374 FAX 492 1071
-