1. Input: Data and sparsity ; learning rate ; the maximum number of iterations ; subset cardinality ; adaptive smoothing factor smoothing vector , and is a constant;
2. Construct the support index set , which includes the index corresponding to the largest in
3. Calculate the sparse initial estimate
where the weighting factor ;
4. Initialize
where is obtained by zero-filling ;
5. fortodo
where is the hard threshold operator, and the gradient
