EPFL_BP all bin - Githubissues

xiyuyi / PyPRIS

This is a python package to perform progressive refinement method for sparse recovery (PRIS)

Other

0 stars 0 forks source link

EPFL_BP all bin #98

Open xiyuyi opened 5 years ago

xiyuyi commented 5 years ago

test EPFL_BP data with all bin

xiyuyi commented 5 years ago

visualizing

xiyuyi commented 5 years ago

chose two alpha, mu values for further calculation.

xiyuyi commented 5 years ago

PyPRIS__Proj_vies_bgSCF2_mu1.0e+08_alpha1.0e-05_pris2_plots_it23001

PyPRIS__Proj_vies_bgSCF2_mu1 0e+08_alpha1 0e-05_pris2_plots_it23001 need to trim Z range

xiyuyi commented 5 years ago

PyPRIS__Proj_vies_bgSCF2_mu5.0e+06_alpha1.0e-05_pris2_plots_it6001

PyPRIS__Proj_vies_bgSCF2_mu5 0e+06_alpha1 0e-05_pris2_plots_it6001

xiyuyi commented 5 years ago

precise results are all bad. now trim PSF

xiyuyi commented 5 years ago

all results were bad. Now try some local test.

xiyuyi commented 5 years ago

now try to use trimmed PSF, but without subtracting the background. local test. Use the same parameter set as bgSCF2_mu5.0e+06_alpha1.0e-05, but with 50000 total steps.

xiyuyi commented 5 years ago

submitted cases:

[xiyuyi@login1 PyPRIS_EPFL_BP_binall_hoffman2_set6]$ pwd
/u/scratch/x/xiyuyi/EPFL_BP/PyPRIS_EPFL_BP_binall_hoffman2_set6
[xiyuyi@login1 PyPRIS_EPFL_BP_binall_hoffman2_set6]$ ls
bgSCF2_mu1.0e+07_alpha1.0e-05_thres-11zrange-60to60
bgSCF2_mu1.0e+07_alpha1.0e-06_thres-11zrange-60to60
bgSCF2_mu1.0e+07_alpha1.0e-07_thres-11zrange-60to60
bgSCF2_mu1.0e+07_alpha5.0e-05_thres-11zrange-60to60
bgSCF2_mu5.0e+06_alpha1.0e-05_thres-11zrange-60to60
bgSCF2_mu5.0e+06_alpha1.0e-06_thres-11zrange-60to60
bgSCF2_mu5.0e+06_alpha4.0e-06_thres-11zrange-60to60
bgSCF2_mu5.0e+06_alpha8.0e-06_thres-11zrange-60to60

xiyuyi commented 5 years ago

trim the PSF would affect the back-projection pattern of the residuals. because errors are back-projected through the transpose of the sensing matrix!

Be careful here.

xiyuyi commented 5 years ago

added more with psf that has the background removed. (subtracting 100)


[xiyuyi@login1 PyPRIS_EPFL_BP_binall_hoffman2_set7]$ pwd
/u/scratch/x/xiyuyi/EPFL_BP/PyPRIS_EPFL_BP_binall_hoffman2_set7
[xiyuyi@login1 PyPRIS_EPFL_BP_binall_hoffman2_set7]$
[xiyuyi@login1 PyPRIS_EPFL_BP_binall_hoffman2_set7]$ ls
bgSCF2_mu5.0e+07_alpha1.0e-07_thres-11zrange-60to60
bgSCF2_mu5.0e+07_alpha1.0e-08_thres-11zrange-60to60
bgSCF2_mu5.0e+07_alpha1.0e-09_thres-11zrange-60to60
bgSCF2_mu5.0e+07_alpha5.0e-08_thres-11zrange-60to60
bgSCF2_mu5.0e+07_alpha5.0e-09_thres-11zrange-60to60

xiyuyi commented 5 years ago

terminated all the jobs. because the local test didn't work.

xiyuyi commented 5 years ago

Now do the following:

[x] reproduce exactly PyPRIS__Proj_vies_bgSCF2_mu1.0e+08_alpha1.0e-05_pris2_plots_it23001. submitted and calculating on Hoffman2. [xiyuyi@login4 PyPRIS_EPFL_BP_binall_hoffman2_set8]$ pwd /u/scratch/x/xiyuyi/EPFL_BP/PyPRIS_EPFL_BP_binall_hoffman2_set8 [xiyuyi@login4 PyPRIS_EPFL_BP_binall_hoffman2_set8]$ ls bgSCF2_mu1.0e+08_alpha1.0e-05_thres-11zrange-60to60 bgSCF2_mu1.0e+08_alpha1.0e-06_thres-11zrange-60to60 bgSCF2_mu1.0e+08_alpha5.0e-06_thres-11zrange-60to60
[ ] Push for publication quality figures.

xiyuyi commented 5 years ago

maybe need some energy filter or something. we'll see.

xiyuyi commented 5 years ago

maybe need to apply mask