This is an example code of the stream-based active learning proposed in our CVPR 2012 paper. The code is provided for research purposes only, with no warrantee. This source code is released under the terms of the GNU Public License. 

External dependencies:
Netlab 3.3: http://www1.aston.ac.uk/eas/research/groups/ncrg/resources/netlab/downloads/

Incremental DPMM learning with memory-bounded variational inference strategy is currently not supported in this version. Our code is now based on the Variational Dirichlet Process Gaussian Mixture Model (VDPGM) written by Kenichi Kurihara
See https://sites.google.com/site/kenichikurihara/academic-software/variational-dirichlet-process-gaussian-mixture-model

The VDPGM package is included in the external folder.

Run the demo.m for a comparison of the proposed active learning method with random sampling.

Set doPlot=1 in globalSetting.m to illustrate the learning process.
Adjust the budget value in globalSetting.m to change the number of supervision budget given. Note that you also need to change activeBudget in demo.m to the same value.

Please cite the following paper if you use this piece of code in your research
Stream-based Joint Exploration-Exploitation Active Learning 
C. C. Loy, T. M. Hospedales, T. Xiang, and S. Gong 
IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2012)


History:
Version 1.00  (09-Jul-2012)
  - Initial version








