Project: jvfeatures

jvtypes.h      jvfeatures.h      chessSeg.cpp      jvtypes.cpp      jvtest.cpp      jvfeatures.cpp     

Project: Other


Project: Infinite HMM Tutorial

run.m      HDP_HMM.m      README.txt      ConditionalProbabilityTable.m      HDP.m      HMMProblem.m      HMM.m     

Project: RRT

RRT.h      RRT.tgz      rrt_test.cpp      RRT.cpp      BidirectionalRRT.cpp      AbstractRRT.cpp     

Project: Box2D_friction_mod

WheelConstraint.h      b2FrictionJoint.h      python_friction_joint.patch      TestEntries.cpp      TopDownCar.h      b2FrictionJoint.cpp      box2d_friction_joint.patch     

Project: Dirichlet Process Mixture Tutorial

EM_GM.m      DP_Demo.m      DPMM.m      DirichletProcess.m      gaussian_EM.m     

Project: Arduino_Code      arduino-serial.c      oscilloscope.pde      motordriver.pde      helicopter_controller.pde      accelerometer_test.pde      ranger_plane_sweep.pde      clodbuster_controller.pde      pwm_manual.pde      ranger_test.pde      servo_test.pde     

Project: ArduCom     

Project: support

geshi.php      Protector.php     

Project: Cogent

CodePane.php      NotesPane.php      PicsPane.php      Cogent.php      PubsTable.php     
Click here to download "resources/code/Infinite HMM Tutorial/README.txt"

resources/code/Infinite HMM Tutorial/README.txt

iHMM Demo Library

This directory contains code for training fixed and variable-sized HMM
models. The implementation and example are based on the infinite Hidden
Markov Model, presented at NIPS in 2003 by Beal et al.  

This is NOT intended for use as an actual sampler for the iHMM - it's
alpha and not remotely game-ready.  It was written as an exercise and to
share it with my lab group at Georgia Tech.  After that meeting I had
little excuse to work out the kinks.  One notable omission is the
resampling of alpha and beta (the HDP concentration params) in the iHMM
example, which I have stubs for but never finished.  

* Also, in a moment of insanity I refactored the code into a fully OOP
implementation, which seems inadvisable in Matlab.  Factoring this way
made the code more readable for me, but it seemed to run far slower than
the imperative version (I blame the GC).  

** If you're looking for a usable, turn-key implementation, I suggest you
check out Jurgen Van Gael's site:

Jonathan Scholz

About me

Pic of me