iron76 / bnt_time_varying

Maximum-a-posteriori dynamic estimation for linkwise dynamic quantities
Other
1 stars 1 forks source link

This is a code to be used with the BNT toolbox for matlab by Kevin Murphy. The original toolbox can be downloaded from (https://code.google.com/p/bnt/). The plugin in this repository modifies the EM estimation procedure
described in www.ai.mit.edu/~murphyk/Papers/learncg.pdf The modifications allow to estimate the variance matrices $\Simga_i$ by assuming that $Bi$ is known but time-varying in the sense that its value is a function of t, i.e. B{i,t} in equation (1).

DEPENDENCIES

Currently the dependencies are (a) BNT toolbox and (b) FEATHERSTONE toolbox: (0.a) https://github.com/bayesnet/bnt.git (0.b) http://royfeatherstone.org/spatial/v2/spatial_v2.zip

DOCUMENTATION

If you want to understand the code in bnt_time_varying, you should understand how to use the BNT and SpatialV2 libraries, that are used in all the code. Nice entry points for this libraries are: