Copyright (C) 2023 Jing Wang

The GRC optimization problem $$\mathop{\min}_{\alpha}||y-X\alpha||_s^s+\lambda||\alpha||_p^p.$$

Four experiments were conducted to compare GRC with LRC, CRC, and SRC on four benchmark face databases including the AR, FEI, FERET, and UMIST face databases.
Experiment 1: Tune the parameters $s$ and $p$ for GRC when principal components that explain 98% of total variance are extracted.
Experiment 2: Compare GRC with LRC, CRC, and SRC when principal components with different percentages (90%, 95%, 98%) of explained variance are extracted.
Experiment 3: Tune the parameters $s$ and $p$ for GRC when different numbers of principal components (54, 120, 200, 300) are extracted.
Experiment 4: Compare GRC with LRC, CRC, and SRC when different numbers of principal components (10:10:300) are extracted.

1. Run load_data.m to split the face databases into training and testing sets.
2. Run the following scripts sequentially:
   1. exp_1_classify.m
   2. exp_1_analysis.m
   3. exp_2_classify.m
   4. exp_2_analysis.m
   5. exp_3_classify.m
   6. exp_3_analysis.m
   7. exp_4_classify.m
   8. exp_4_analysis.m
3. More benchmark face databases are refered to: https://github.com/yuzhounh/Face_databases.

Jing Wang
[email protected]
2023-11-23 10:19:28