Pointers to some code which students have made available for some projects.
No guarantee or support of any kind of course...
- Continuous Profile Model for Time Series Alignment
Jennifer Listgarten has posted some code
for the CPM model here.
- Modeling Human Motion Using Binary Latent Variables
Graham Taylor has posted some code and data
for the our mocap model here.
- Kurtotic ICA
A two line version of ICA!.
[version of dec.01-->kica page]
- Hidden Markov Models
Trivial MATLAB code for HMMs with both discrete
and continuous observations, including scaling tricks and support for
[version of aug.99-->tar.gz(13K)]
- EM algorithms for PCA/SPCA
Here is the MATLAB code to implement this algorithm.
[version of june 98 (code only)-->
[version of june 98 (code and toy data)-->
- Linear Dynamical Systems
Maybe eventually I'll post MATLAB code for LDSs (Kalman filter models)
including the EM algorithm for parameter training. But for now, see Zoubin's code linked below.
[version of aug.99-->tar.gz(??K)]
Byte Boundaries on Suns/x86
- Chris Harvey wrote this small piece of C code to check
if your architecture will allow you to write a double that isn't
exactly aligned with a even byte boundary. Suns won't let you do this, x86
- Extracting Data from CVS Repositories in SQL
Python code written by our summer research assistant, Keir Mierle
to process CVS repositories (both the RCS files and the master CVS
history log) and load the transaction information into a SQL database.
- Wisconsin X-ray Microbeam Data Analysis
Hopefully I can find the time to clean up and post MATLAB code for reading/writing and analysis
of the Wisconsin x-ray microbeam speech database (UBDB). This database
contains simultaneous speech and articulator movement information.
See my UBDB page for more info.
MATLAB code for dowloading and parsing the data output by
the Crosspad digital writer product from Cross/IBM.
See my Crosspad page for more info.
External Code Links
- Kevin Murphy (General Bayes Nets)
Both HMMs and LDSs are examples of directed acyclic graphical
models (in fact they have the same graphical model, just discrete
state nodes for HMMs and continuous state for LDSs).
Kevin Murphy's excellent code packages do learning and inference
for almost any directed graphical model with mixed continuous and
discrete nodes. (This means his code basically does all that my code
does and more.)
In fact, you should check out all of Kevin's page because he does
interesting and very good work.
- Carl Rasmussen (conjugate gradient and kernel density)
Carl Rasmussen (now at Max Planck Tuebingen)
small page which includes code for conjugate gradient
and kernel density estimation.
- Zoubin Ghahramani (HMM,LDS&others)
Zoubin also has a nice
software page which includes code packages to do
learning and inference in Hidden Markov Models, Linear Dynamical
Systems and other more sophisticated models.
Sam Roweis, Vision, Learning and Graphics Group,