This library provides the MIL forward log-likelihood of single-channel data given a hidden Markov model.
It also has subinterval likelihood (http://www.qub.buffalo.edu/wiki/index.php/Maximum_Subinterval_Likelihood)
and parameterization with linear constraints.

max_ll_util contains utilities common to both algorithms.  It exports functions to set up a system
of constrained parameters.  Inside the library it provides common functions like missed event correction
and equilibrium probability.

max_inter_ll contains the MIL forward LL along with functions to prepare the idealized data.

max_subi_ll is a variation on MIL for stimulated data -- it allows additional idealized signals to control
model variables such as Ligand and Voltage.  It's called sub-interval likelihood because, when the
stimulus changes mid-dwell, each partial dwell must be considered separately.

The library should compile for Windows (VC7.1) and Linux (make).  You'll need to
get a recent boost distribution (I used 1_39_0) and make sure it's in the include path.

Hopefully instead of compiling, you can just use the 32-bit dll provided,
along with max_inter_ll.h (c/c++) or maxill_iface.pas and maxill_utils.pas (delphi).

max_inter_ll.py and max_subi_ll.py have the same programs, written to python/numpy/scipy.
They also test maxill.dll [libmaxill.so] if available.  Despite numpy, they are
much slower due to repeated allocation/deallocation.