Active Discriminative Dictionary Learning for Weather Recognition

Algorithm 1

Active discriminative dictionary learning (ADDL).

Inputs: Labeled set and its label set , Unlabeled set , the number of iteration

    and the number of unlabeled samples to be selected in each iteration.

  (1) Initialization: Learn an initial dictionary pair and by DPL algorithm from the .

  (2) For   to , do

  (3) Compute and by (6) and (10) for each sample in the unlabeled dataset .

  (4) Select samples (denoted by ) from the by (13), and add them into with their

     class labels which manually assigned by user. Then updates and .

  (5) Learn the refined dictionaries and over the expanded dataset .

  (6) End for

Output: Final learned dictionary pair and .