problems with optimization and convergence
47 views
Skip to first unread message
Lena Zemlemerova
Dec 5, 2024, 8:46:19 AM12/5/24
to dadi-user
Hello, dadi-users.
I tried to find parameters of optimization for two-population models, using dadi_Run_Optimizations.py according Portik et al., 2017. But I couldn't find the parameters that lead models to convergence. I tried to increase the number of rounds but it didn't improve the picture of convergence (see pict 1,2+file1). I tried to add different boundaries but it didn't solve my problem (pic 3-6, files 2,3). There are the parameters of optimization for "sec_contact_sym_mig_size" model. I have two populations of 34 and 24 samples and tried to determine the time oftheir second contact.
I tried to find parameters of optimization for two-population models, using dadi_Run_Optimizations.py according Portik et al., 2017. But I couldn't find the parameters that lead models to convergence. I tried to increase the number of rounds but it didn't improve the picture of convergence (see pict 1,2+file1). I tried to add different boundaries but it didn't solve my problem (pic 3-6, files 2,3). There are the parameters of optimization for "sec_contact_sym_mig_size" model. I have two populations of 34 and 24 samples and tried to determine the time oftheir second contact.
For the first case (see pict 1,2+file1)
pts = [100,110,120]
p_labels = "nu1a, nu2a, nu1b, nu2b, m, T1, T2"
reps = [10,20,30,30,30,30,30,30,30,30]
maxiters = [3,5,10,15,15,15,15,15,15,15]
folds = [3,2,2,1,1,1,1,1,1,1]
p_labels = "nu1a, nu2a, nu1b, nu2b, m, T1, T2"
reps = [10,20,30,30,30,30,30,30,30,30]
maxiters = [3,5,10,15,15,15,15,15,15,15]
folds = [3,2,2,1,1,1,1,1,1,1]
the second case (see pict 3,4+file2)
pts = [100,110,120]
p_labels = "nu1a, nu2a, nu1b, nu2b, m, T1, T2"
upper = [100,100,100,100,200,10,10]
lower = [1e-3,1e-3,1e-3,1e-3,1e-5,1e-3,1e-3]
reps = [10,20,30,30,30,30,30,30,30,30]
maxiters = [3,5,10,15,15,15,15,15,15,15]
folds = [3,2,2,1,1,1,1,1,1,1]
p_labels = "nu1a, nu2a, nu1b, nu2b, m, T1, T2"
upper = [100,100,100,100,200,10,10]
lower = [1e-3,1e-3,1e-3,1e-3,1e-5,1e-3,1e-3]
reps = [10,20,30,30,30,30,30,30,30,30]
maxiters = [3,5,10,15,15,15,15,15,15,15]
folds = [3,2,2,1,1,1,1,1,1,1]
the third case (see pict 5,6+file3)
pts = [100,110,120]
p_labels = "nu1a, nu2a, nu1b, nu2b, m, T1, T2"
upper = [100,100,100,100,10,20,20]
lower = [0.01,0.01,0.01,0.01,0,0,0]
reps = [10,20,30,30,30,30,30,30,30,30]
maxiters = [3,5,10,15,15,15,15,15,15,15]
folds = [3,2,2,1,1,1,1,1,1,1]
p_labels = "nu1a, nu2a, nu1b, nu2b, m, T1, T2"
upper = [100,100,100,100,10,20,20]
lower = [0.01,0.01,0.01,0.01,0,0,0]
reps = [10,20,30,30,30,30,30,30,30,30]
maxiters = [3,5,10,15,15,15,15,15,15,15]
folds = [3,2,2,1,1,1,1,1,1,1]
Appreciate all the advice!
Elena
Ryan Gutenkunst
Dec 16, 2024, 2:27:42 PM12/16/24
to dadi-user
Hello Lena,
Sorry for the slow reply. I’m not super familiar with the Portik pipeline, so it’s difficult for me to interpret these results. Have you tried starting with simpler models?
You could also try using dadi-cli, our command-line interface for dadi which provides a lot of quality-of-life improvements to the user experience: https://github.com/xin-huang/dadi-cli
Best,
Ryan
--
You received this message because you are subscribed to the Google Groups "dadi-user" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dadi-user+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/dadi-user/7ccbb278-41ed-44df-9894-b9c20db637ben%40googlegroups.com.
<fig3.jpg><fig1.jpg><fig2.jpg><File1.txt><fig5.jpg><File3.txt><fig4.jpg><fig6.jpg><File2.txt>
Reply all
Reply to author
Forward
0 new messages