how can I save the verbose output from fitrgp?

I am running fitrgp with optimization options as following:

gprMd3 = fitrgp(x,y,'KernelFunction','squaredexponential',...
    'OptimizeHyperparameters','auto','HyperparameterOptimizationOptions',...
    struct('AcquisitionFunctionName','expected-improvement-plus'));

and I get the following table with two plots: one plot for the objective function model, and other plot for Min objective vs. Number of function evaluations

The table I get is:

|======================================================================================|

| Iter | Eval | Objective: | Objective | BestSoFar | BestSoFar | Sigma |

| | result | log(1+loss) | runtime | (observed) | (estim.) | |

|======================================================================================|

| 1 | Best | 0.0023244 | 83.433 | 0.0023244 | 0.0023244 | 0.00013322 |

| 2 | Accept | 0.009627 | 38.782 | 0.0023244 | 0.0027615 | 0.15882 |

| 3 | Accept | 0.0025183 | 56.103 | 0.0023244 | 0.004802 | 0.01944 |

| 4 | Accept | 0.0024726 | 79.114 | 0.0023244 | 0.0042355 | 0.0017215 |

| 5 | Accept | 0.0025118 | 76.332 | 0.0023244 | 0.0023272 | 0.00014478 |

I would like to save this numbers or access to them. So that I can reproduce the two plots above.

Is there a way to do it?

Thank you!

Answer 1

[Pratik]

Hi Franisco,

From what I can understand, the verbose output from ‘fitgrp’ must be saved so that the plots can be reproduced.

In MATLAB, you can capture the verbose output of the ‘fitrgp’ function's hyperparameter optimization by setting the Verbose option to 1 or 2 in the ‘HyperparameterOptimizationOptions’ structure. Additionally, to save the detailed optimization information, you can set the ‘SaveIntermediateResults’ field to true. This will allow you to access the optimization history, including all the intermediate results.

Please refer to the table in the following documentation for more information:

www.mathworks.com/help/stats/fitrgp.html#butnn96_sep_shared-HyperparameterOptimizationOptions

I hope this helps!