Jacobian issue resolved
The discrepancy mentioned in my previous blog post is resolved. The Jacobian calculation formula was working fine. It turned out some code debugging was required. Now the jacobian is calculated explicitly for each step using forward differences method. It is then passed on to leastsq function for the calculation of parameters.
Next I will build some mechanism to let the user pass his own jacobian calculating function.
The code edits can be viewed here.
Viewing parameter estimates
The explicit jacobian approximations helps to save the parameter estimates at each iteration of the LM algorithm. These parameter estimates are passed on to an instance of the results class.
Presently the user can view the estimates using view_iter() function of the result instance. A screenshot of the parameter estimates table is given below.
The discrepancy mentioned in my previous blog post is resolved. The Jacobian calculation formula was working fine. It turned out some code debugging was required. Now the jacobian is calculated explicitly for each step using forward differences method. It is then passed on to leastsq function for the calculation of parameters.
Next I will build some mechanism to let the user pass his own jacobian calculating function.
The code edits can be viewed here.
Viewing parameter estimates
The explicit jacobian approximations helps to save the parameter estimates at each iteration of the LM algorithm. These parameter estimates are passed on to an instance of the results class.
Presently the user can view the estimates using view_iter() function of the result instance. A screenshot of the parameter estimates table is given below.
No comments:
Post a Comment