Robust Matching Pursuit Extreme Learning Machines

Algorithm 1

KRSLMP-ELM.

Input: samples

Output: weight vector

  Parameters setting: number of hidden nodes , regularization parameter and sparsity level .

  Initialization: randomly initialize ELM parameters: input weights and biases in measurement matrix .

   Set the index set , the residual , the iteration counter and .

 for      do

   Find a column of most correlated with the residual

   Augment the index set

   Solve the KRSLMP minimization problem by the following iterations

     The solution is denoted as

   Update residual

  end   for