b'\n \n \n
 
\n  
sfvelinvnew (4.0)
index
user/seisinv/Mvelinvnew.f90
\n None\n

\n \n \n \n \n \n
 
\n Synopsis
       sfvelinvnew < infile.rsf res=fres.rsf vel0=fm.rsf mres=fmres.rsf > outfile.rsf velout=vtr.rsf nv=nhx dv=0.01 ov=1.5 niter=20 savevel=0 flag=0 mflag=0 huber=0 irls=0 nstep=1 rwt=0. mwt=0. srate=0.01 eps=0.01 lamda=1000. delta=0.0001 step=0.000005 alpha=790.635 alpha= lip= reset=
\n

\n \n \n \n \n \n
 
\n Parameters
       \n \n \n
\n  
alpha=
\tsmoothing parameter, typical value: 1 to 10 times estimated norm(x,inf)
\n
\n \n\n \n \n
\n  
delta=0.0001
\tdelta controls update step and convergent, small delta ensure convergence but with small decrease in data fit error
\n
\n \n\n \n \n
\n  
dv=0.01
\t
\n
\n \n\n \n \n
\n  
eps=0.01
\t
\n
\n \n\n \n \n
\n  
flag=0
\t
\n
\n \n\n \n \n
\n  
huber=0
\t
\n
\n \n\n \n \n
\n  
irls=0
\t
\n
\n \n\n \n \n
\n  
lamda=1000.
\tlamda controls sparsity, bigger lamda, more sparsity
\n
\n \n\n \n \n
\n  
lip=
\tthe estimated Lipschitz constrant of the dual objective, default: alpha*normest(A*A\',1e-2)
\n
\n \n\n \n \n
\n  
mflag=0
\t
\n
\n \n\n \n \n
\n  
file mres=
\tauxiliary output file name
\n
\n \n\n \n \n
\n  
mwt=0.
\t
\n
\n \n\n \n \n
\n  
niter=20
\t
\n
\n \n\n \n \n
\n  
nstep=1
\t
\n
\n \n\n \n \n
\n  
nv=nhx
\t
\n
\n \n\n \n \n
\n  
ov=1.5
\t
\n
\n \n\n \n \n
\n  
file res=
\tauxiliary output file name
\n
\n \n\n \n \n
\n  
reset=
\tNesterov\'s acceleration restart (theta is reset) or skip (theta is not reset)
\n
\n \n\n \n \n
\n  
rwt=0.
\t
\n
\n \n\n \n \n
\n  
savevel=0
\tFlag to choose the algorithm
\n
\n \n\n \n \n
\n  
srate=0.01
\t
\n
\n \n\n \n \n
\n  
step=0.000005
\tstep is very important in convergence and sparsity
\n
\n \n\n \n \n
\n  
file vel0=
\tauxiliary input file name
\n
\n \n\n \n \n
\n  
file velout=
\tauxiliary output file name
\n
\n \n
'