##Author: Yao-Hung (Hubert) Tsai [email protected]
#####Package with code and demo usage for the paper:
#####"Learning Cross-Domain Landmarks for Heterogeneous Domain Adaptation"
- Dowload and compile libsvm with weights support
- The path to libsvm-weights code available at https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/weights/libsvm-weights-3.20.zip
- Edit CDLS_demo.m with desired parameters and path
- edit the path to /libsvm-weights/matlab
- edit the paramters of iter, delta, and PCA_dimension if you want
- Put you own data in /data
- E.g., /data/amazon_DeCAF_dslr_SURF.mat is a random split of images from Office and Caltech-256 Datasets (Amazon images with DeCAF features and DSLR images with SURF features)
- Directly run CDLS_demo.m