Preannounce and vote on the CRM114 library...

  [ sorry for the long drought of no messages ]

Yes, it's now been done... over half of the CRM114 classifiers have 
now been ported to a library format, with a C-friendly interface
(no need to worry about the language's overlapping strings stuff).

Currently the library supports OSB/Markov, Hyperspace, FSCM, SVM, and
Bit Entropy classifiers, with Neural Net, OSBF, and corellator
filters "to be done soon".  Yes, it's still a work-in-progress.

It's also embedded-system friendly - it still presumes a stderr, but
needs file system; classifier data files are now just single blocks of 
memory.  It's faster, and cleaner, and we fixed a bunch of 
trivial little bugs in the process of the librarification.

This is ONLY the classifiers (i.e. the LEARN and CLASSIFY
commands).  You call it from C, or any language that can call
C.  There's no need to learn a new language.

The size is much, MUCH smaller than full CRM114.  It won't fit on 
a PIC chip, but it will definitely fit on a Via C7 or other 
such mini-sized X86.  It might even fit on a GumStix; the 
X86-64 library is about 400Kbytes, and most of that 
is the SVM code.   However, a WRT54G is probably a little too
small (insufficient space to store the classifier tables), and
it probably won't fit on an FPGA either.

Typical footprint is 400K for the code, and 8 to 16 megabytes
per class, so a machine with 256 megabytes is perfectly adequate.

To use the library, you will also need a regex lib (TRE is 
default, and TRE is BSD-licensed), and a C compiler (GCC is
recommended).

The CRM114 library itself is LGPLed.  We haven't backported
the library of classifiers to the CRM114 engine yet.

   - Now, the question -

The CRM114 library is currently prefixed as "libcrm", as in
"libcrm_learn_text(...)".  But if we should change it, now's the
time, as once we release, we really ought to keep this stable.

Do people prefer:

   libcrm_... ?  ... which is what's there now.

   libcrm114_... ?  (avoids confusion with other CRMs like SugarCRM, which
   	     	is a Customer Relationship Management system)

   crm114_... ? makes it clear what the library is, but doesn't 
                make it clear that this is NOT the language.

Wisdom, please?

(and yes, releasing the package and docs and demo set is on the
to-do list!)

      - Bill Yerazunis

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev