Improved calibration strategy for LeastSquares #450
raoulcollenteur
started this conversation in
Ideas
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi All,
Here's a more advanced parameter estimation strategy that helped me lately.
ml.solve(tmin=tmin, tmax=tmax, x_scale="jac", jac="3-point", tr_solver="exact")
what this does is 1) scale the steps that are taken when changing the parameters between the iterations by the Jacobian, and 2) evaluate the Jacobian using a more accurate 3-point rather than the default 2-point scheme. Scipy LeastSquares options described in more detail here: https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.least_squares.html
I found that this improved the parameter estimation particularly for more complex models (more params), making it more robust. The downside is that it is slower, but who cares?
Perhaps others can also test this.
Cheers,
Raoul
Beta Was this translation helpful? Give feedback.
All reactions