Uncorrelated Multilinear Discriminant Analysis (UMLDA)

The codes implement the Uncorrelated Multilinear Discriminant Analysis (UMLDA) algorithm.

Matlab source codes for Uncorrelated Multilinear Discriminant Analysis (UMLDA)

%[Algorithm]%

The matlab codes provided here implement the UMLDA algorithm (as well as its
regularized and aggregated versions) presented in the paper "UMLDA_TNN09.pdf"
included in this package:

Haiping Lu, K.N. Plataniotis, and A.N. Venetsanopoulos,
"Uncorrelated Multilinear Discriminant Analysis with Regularization and Aggregation for
Tensor Object Recognition",
IEEE Transactions on Neural Networks,
Vol. 20, No. 1, Page: 103-123, Jan. 2009.

[Files]
RUMLDA.m: the Regularized UMLDA (R-UMLDA)
demoRUMLDAAggr.m: sample code for R-UMLDA aggregation with sample output
estMaxSWEV.m: estimate \lambda_{max} in the paper, used for regularization
---------------------------

%[Data]%

All data used in the paper are included in this package:

Directory "PIEP3I3" contains the PIE face data and their partitions used in the paper.
Directory "FERETC80A45S6" contains the FERET face data for C=80 and their partitions.
Directory "FERETC160A45S6" contains the FERET face data for C=160 and their partitions.
Directory "FERETC240A45S6" contains the FERET face data for C=240 and their partitions.
Directory "FERETC320A45S6" contains the FERET face data for C=320 and their partitions.
Directory "USFGait17_32x22x10" contains the gait data used in the paper.
---------------------------

%[Usages]%

Please refer to "demoRUMLDAAggr.m" for example usage on 2D data
"FERETC80A45S6_32x32" in the directory "FERETC80A45S6", which is used in the
paper above. The partition used in the paper is included in the directory
"FERETC80A45S6\4Train" for L=4.
---------------------------

%[Sample face recognition results for reference]%

calcR1.m: calculate the classification rates for aggregated learners

Run demoRUMLDAAggr.m to get sample face recognition results

FRSampleOutput.txt contains sample output* in the command window.
*Note: The results won't be identical because random initialization is involved.
However, the deviation should be small (around 2%).
---------------------------

%[Toolbox needed]%:

This code needs the tensor toolbox available at
http://csmr.ca.sandia.gov/~tgkolda/TensorToolbox/
This package includes tensor toolbox version 2.1 for convenience.
---------------------------

%[Restriction]%

In all documents and papers reporting research work that uses the matlab codes
provided here, the respective author(s) must reference the following paper:

[1] Haiping Lu, K.N. Plataniotis, and A.N. Venetsanopoulos,
"Uncorrelated Multilinear Discriminant Analysis with Regularization and Aggregation for
Tensor Object Recognition",
IEEE Transactions on Neural Networks,
Vol. 20, No. 1, Page: 103-123, Jan. 2009.
---------------------------

%[Additional Resources]%

The BibTeX file "UMLDApublications" contains the BibTex for UMLDA and
related works. The included survey paper "SurveyMSL_PR2011.pdf" discusses the
relations between UMLDA and related works.