The Daily Static
  The Daily Static
UF Archives
Register
UF Membership
Ad Free Site
Postcards
Community

Geekfinder
UFie Gear
Advertise on UF

Forum Rules
& FAQ


Username

Password


Create a New Account

 
 

Back to UserFriendly Strip Comments Index

About car efficiency... (and re: Skippy and Tyop) by etwas_egal2007-04-30 00:07:38
  Consolidate errands by shadowboy8132007-04-30 06:06:51
    Refinement to point 6: *accelerate* slower. by etwas_egal2007-04-30 07:46:13
      No, I meant *DRIVE* slower for point 6 by shadowboy8132007-04-30 10:12:30
        Highway speed limits here are at your break point by etwas_egal2007-04-30 11:34:10
          Any clustering I'm currently doing... by shadowboy8132007-04-30 14:20:21
            I figured you would be doing something like that by etwas_egal 2007-04-30 14:54:51
based on your field. I currently have 1500-space and 40 million records (the summary version of about two billion records). But I *know* I have a latent structure in what I hope is a lower-dimension space.

I'm looking at SparseLib++ with sparse-BLAS among other things, since most of the records have a substantial number of zeros (or missing :( ). Thought I'd check to see if there was another good package.

I'm probably going to be starting with a simple principle components and support vectors, and see if that reveals enough to get a better idea going from there. But in any case, I'm having a hard time coming up with a meaningful measure of "goodness-of-dimensionality" since percent of variance explained just doesn't tell me if I have retained the power I will need in the final structure. *sigh*
[ Reply ]
              Well, as you can imagine, my matrices are dense by shadowboy8132007-04-30 16:55:25

 

[Todays Cartoon Discussion] [News Index]

Come get yer ARS (Account Registration System) Source Code here!
All images, characters, content and text are copyrighted and trademarks of J.D. Frazer except where other ownership applies. Don't do bad things, we have lawyers.
UserFriendly.Org and its operators are not liable for comments or content posted by its visitors, and will cheerfully assist the lawful authorities in hunting down script-kiddies, spammers and other net scum. And if you're really bad, we'll call your mom. (We're not kidding, we've done it before.)