#PLGM means Pairwise Linear Gaussian Model.
Programs written by valerian.nemesin@gmail.com
README.markdown written by stephane.derrode@ec-lyon.fr
Remarks :
-
Papers explaining the algorithms can be found in Valérian Némesin and in Stéphane Derrode web pages (or here).
-
Note that a demo of PLGM algorithms still works (how long ?) to test the EM algorithm (without constraint). Don't be afraid if the figures are missing, you can still download the generated and estimated data using the link! Remark At the time of publication the Web site of Institut Fresnel (where the demo is hosted) is not working, but we can expect that to be a temporal problem and all went OK in a few hours or days.
-
A new front-end for Python and Matlab is being written and should appear in the next week to easy the parametrization of Partial learning configuration files.
##1. Compile all programs
-
From the root repository of the project, change options in
./compilation.mk
and in./tkalman_c/Applications/options.mk
to suit your needs, especially la variableAPP_DIR
must indicate where applications will be stored (but default should be OK). -
Install gsl library and
pkg-config
. For GSL you need thedev
library (meaning that you should also install GSL headers files for compilation). -
On Ubuntu, you need to install the package
xutils-dev
for the commandmakedepend
used by the makefiles. Install it usingsudo apt install xutils-dev
. -
Then
make
, ormake forced
. If all went well, the programs to run should be inAPP_DIR
.
##2. Test Kalman programs with Octave scripts
-
Go to
Octave
repository and runoctave
(should be installed). -
Open one experiment from
exp_*.m
files (start withexp_couple.m
for example), and change option according to your needs. Then run. A repository calledResultats
should hold the results, with files and figures in png format.
Good luck!