Open LinaMiao opened 12 years ago
-- works fine for help_spgl1 example. which can be seen from the publish file in task 16(help_spgl1)
-- next try on simu shots and l1 migration problem.
-- for convolution problem, generally decrease the residual faster. Both of those two situations, either method can give right amplitudes
-- for PQN if given more iterations, updatetau every iteration, the result is not so bad, but this increase the function evaluation time significantly.
Needs to be fixed.
See script2.pdf
-- modify output info, correct time and add a total projects in
-- reflect user defined optTol in Lasso -- based on code from task 15 -- make sure each lasso problem being terminated correctly