search

jlapeyre / PDL-Fit-Levmar

PDL interface to constrained Levenberg-Marquardt library levmar
Other
3 stars 3 forks source link

PDL-Fit-Levmar-0.0103 has started to fail #7

Closed andk closed 2 years ago

andk commented 2 years ago

As per subject. Sample fail report: http://www.cpantesters.org/cpan/report/ee71abd4-7840-11ec-aff9-28db217d911e

Statistical analysis seems to indicate it has started with PDL-2.064:

Regression 'mod:PDL'

Name Theta StdErr T-stat [0='const'] 1.0000 0.0000 40864836807746921320.00 [1='eq_2.058'] -0.0000 0.0000 -1.17 [2='eq_2.063'] 0.0000 0.0000 1.04 [3='eq_2.064'] -1.0000 0.0000 -28895803218839475540.00 [4='eq_2.065'] -1.0000 0.0000 -28895803218839475540.00 [5='eq_2.066'] -1.0000 0.0000 -28895803218839475540.00 [6='eq_2.068'] -1.0000 0.0000 -28895803218839475540.00 [7='eq_2.072'] -1.0000 0.0000 -28895803218839475540.00 [8='eq_2.074'] -1.0000 0.0000 -28895803218839475540.00

R^2= 1.000, N= 15, K= 9

mohawk2 commented 2 years ago

Thanks for the report! In fact it will be 2.063_04 that did it with xlinvals et al always returning at least a double, despite the tests' best effort to make it be a float, but that can be fixed.

2.073 has also introduce different breakage due to the wrapping code passing in various null as input (now forbidden) and [t] params (now forbidden due to multi-threading), which I am now fixing.

mohawk2 commented 2 years ago

Fixed and released as 0.0104!