[lug] Weighting of factors and whatnot (a bit OT)

Chris Riddoch riddochc at gmail.com
Tue Jan 30 18:27:21 MST 2007


A collected reply:

Evelyn:  Yes, the moment you said it, I had a "Well, *duh*." moment. I
totally missed the similarity of this problem to one I already knew
about - latent semantic analysis, which uses SVD as its primary
mechanism.  The fact that I learned about this at the university and
yet still don't fully grok the math is frustrating.  For anyone
curious:

http://en.wikipedia.org/wiki/Latent_semantic_analysis

I'm still trying to wrap my head around the linear algebra involved,
but I have a hard time with math, and haven't really dedicated enough
brainpower to it.

D. Stimits:  That's an interesting comparison.  I hadn't thought about the

Steve: I was thinking something about vector space. I had something a
little like a cosine similarity in that 6-dimensional space in mind.
I've read about support vector machines in my NLP classes, and have an
abstract understanding of how they work for classification without
really understanding the math.  A kernel, to me, is a piece of
software.  :)

I remain, sadly, just a little out of my depth.

But I really appreciate everyone's help!  Anybody got recommendations
for books like "Really advanced math for people who grok calculus but
suck at algebra"?

And if anyone wants a hint at the data I'm using this for, have a look
at http://online.onetcenter.org/  Let's just say it's completely
selfish motivations.

-- 
epistemological humility
  Chris Riddoch



More information about the LUG mailing list