About the "Fine-tuning on Pre-trained FCN with DSR" in your paper

[yang502]

In your paper, the parameters of FCN and W (sparse coefficient matrix) seems are alternate update. I want to where is the corresponding optimization process implemented? @lingxiao-he

[upgirlnana]

i learn something about LR，i don‘’t understand how the W (sparse coefficient matrix) is learnt. hope your explanation.thank you