home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cs.utexas.edu!sun-barr!news2me.ebay.sun.com!exodus.Eng.Sun.COM!appserv.Eng.Sun.COM!sun!amdcad!kpc!hollasch
- From: hollasch@kpc.com (Steve Hollasch)
- Newsgroups: comp.graphics.visualization
- Subject: Re: Representing hyperplanes ((N-1)-flats in N-space) in 2D (was Re: Neural Networks and HyperPlanes)
- Summary: Representation via successive projections
- Message-ID: <1992Jul22.183709.20068@kpc.com>
- Date: 22 Jul 92 18:37:09 GMT
- References: <1992Jul20.173438.12081@cis.uab.edu> <1992Jul22.033101.24909@iti.gov.sg>
- Sender: usenet@kpc.com
- Organization: Kubota Pacific Computer, Inc.
- Lines: 30
-
- vachha@cisa.cis.uab.edu (Rustom (Yuppy) Vachha) writes:
- > Hence we need to define a strategy to convert N-dimensional data into 2D.
- > How do we represent N dim hyperplanes?
-
- My thesis on this topic used two approaches. The raytracing approach
- fires rays through a 3D grid and yields a 3D image (3D RGB field); I don't
- think this is what you're looking for.
-
- The other approach was for a wireframe viewer. This is not a new
- technique, and basically involves successive projections, e.g. 4D -> 3D,
- and then 3D -> 2D. While you _can_ bundle these projection mathematically
- to 4D -> 2D, the successive nature of the projections allows you to tailor
- parameters for each projection. For example, you can use parallel
- projection for 4D -> 3D, and then perspective projection for 3D -> 2D. Or
- parallel/parallel, perspective/perspective, or perspective/parallel. In
- addition, you can also mess with viewing angles and the like.
-
- This is rather trivial for transforming points, and so is each for
- wireframe data. However, visualization of 4D wireframes is very limited
- in its usefulness. Other approaches have used a scanline (scanplane?)
- approach, but my impression is that this approach would be very complex.
-
- With respect to tiling 4D functional fields, I have thought about how
- you'd tesselate the hyperplane and then view it, but didn't get very far
- in implementing it. I think that this approach might be more confusing
- than a 3D RGB field.
-
- _____________________________________________________________________________
- Steve Hollasch Kubota Pacific Computer, Inc.
- hollasch@kpc.com Santa Clara, California
-