The full text of the article "Bookish Math" may be found at the science
news website
http://www.sciencenews.org
of particular interest to me are support-vector machines.
from the article ....
The problem with Support Vector Machines (SVM's) is that they can take an
incredibly long period of time to train. Bayesian takes comparatively no
time to train. SVM's can take from minutes to hours to days to train
depending on the number of features you train it on. (A feature is a item
that you classify on such as the phrase "Make Money Fast" or "this is not
spam.") The more features and examples, the longer it takes to train SVMs
as you might expect. The problem is that when the time increases, it
doesn't increase in a nice linear fashion as one might hope.
SVMs are great at classification. The only thing is that you need to have
the horsepower and/or time to train them.
-Art
--
Art Pollard
http://www.lextek.com/
Suppliers of High Performance Text Retrieval Engines.
_______________________________________________
Asrg mailing list
Asrg(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/asrg