[staging] 🐛 Optimization does not work with dataset scale parameters
jsnel opened this issue · 1 comments
jsnel commented
Version information
- pyglotaran version: staging
Describe the bug
An optimization involving an experiment with multiple datasets and scaling parameters (of which (n-1) should be estimated) does not currently optimize correctly.
To Reproduce
Steps to reproduce the behavior:
Run sim_analysis_script_3d_disp.ipynb (e.g. here)
Observed behavior
Optimization
Iteration Total nfev Cost Cost reduction Step norm Optimality
0 1 2.8053e+06 5.62e+06
1 2 1.5353e+06 1.27e+06 4.74e+00 3.15e+06
2 3 6.2623e+05 9.09e+05 4.74e+00 2.38e+06
3 5 3.2530e+05 3.01e+05 1.19e+00 2.34e+05
The maximum number of function evaluations is exceeded.
Function evaluations 5, initial cost 2.8053e+06, final cost 3.2530e+05, first-order optimality 2.34e+05.
First plot
Expected behavior
Iteration Total nfev Cost Cost reduction Step norm Optimality
0 1 8.2822e+00 8.54e+03
1 2 8.0162e-05 8.28e+00 3.09e-03 9.46e+00
2 3 7.3664e-05 6.50e-06 2.23e-05 1.67e-04
3 4 7.3664e-05 0.00e+00 0.00e+00 1.67e-04
`xtol` termination condition is satisfied.
Function evaluations 4, initial cost 8.2822e+00, final cost 7.3664e-05, first-order optimality 1.67e-04.
Expected result (first plot_
Extra information
A PR ( #1461 ) was created which fixes the issues, but there may be a better way.