fast_newuoa

Runs first few steps of a NEWUOA optimisation for some synthetic data. This test exists only to increase test coverage of the routines involved in a NEWUOA optimisation (some of which is used by other optimisation routines). It should not be the basis of any scientific runs. The tolerances are all very loose to make each iteration fast enough that the test completes within a few minutes. The data is taken from the output of a 1 \({\rm M}_\odot\) model at an age around 1 Gyr.

To increase coverage, this test uses the custom variables implemented in my_var1, my_var2 and my_var3. Specifically, it sets these to the luminosity, effective temperature and radius of the Sun. It also sets chi2_seismo_fraction to zero, so the test mimics a traditional solar calibration. It almost certainly won’t converge, however, because NEWUOA is prone to rapidly getting stuck in local minima.

The output in effect only checks that at least one set of trial parameters returned a value of \(\chi^2\). There are two main failure modes. First, failure might benignly indicate that the initial parameters have strayed too far from the synthetic constraints to produce any output, in which case the initial guesses or target data should be adjusted. Second, failure might indicate that the NEWUOA optimisation is genuinely broken and needs fixing.

Last-Update: 2022-03-29 (mesa 998d243) by Warrick Ball