Guided Subspace Learning(GSL demo)

This is the source code of the GSL method, which has been published as “Guide Subspace Learning for Unsupervised Domain Adaptation“ IEEE Trans.NNLS, 2020.

In the file, there are multiple .m files.

GSL.m : Core codes of GSL algorithm.

NGSL.m : Core codes of NGSL algorithm.

demoGSL.m : Evaluate GSL/NGSL on an example task (C-W in 4DA dataset).

You can run the "demoGSL.m" code for your reference. Then you will get the results:

GSL: Choose 'primal' as 'kernel_type', the final accrucy will be ‘55.93%’.

NGSL: Choose 'linear'(linear kernel in our paper,for example) as 'kernel_type' , the final accrucy will be ‘63.39%’.

When you use our code, there are two(GSL)/three(NGSL) main parameters need to be adjusted according to different tasks:

alpha,beta and lambda

Once you run the code, please correctly set the path of the data and liblinear toolbox

Please contact leizhang@cqu.edu.cn or jrfu@cqu.edu.cn if there is any problem

Enjoy it!