Added a n_random parameter to smart_start to allow multiple different undeterministic (suboptimal) solutions

in scipy > 1.5.0, slsqp does not have the previous flaw wrt. limits, and applying the fix must now be disabled to make it work

improved post penalty configuration

problems example elaborated. its noted however that make_notebooks is not working due to a bug in the drivers.ipynb

Witold Skrzypiński authored 5 years ago and Mikkel Friis-Møller committed 5 years ago

Inputs for calculate_devex and calculate_capex converted from lists to np arrays within the methods.

default True objective set as False for electrical cost component & x and y changed to topfarm.x_key etc

Emre Barlas authored 5 years ago and Mikkel Friis-Møller committed 5 years ago

added default values to main() to allow pytest to run.

Mikkel Friis-Møller authored 5 years ago and Riccardo Riva committed 5 years ago

New PyWake syntax
Updated docker image with texlive-latex-extra for PyWake report

removed version from init

types as input

Mikkel Friis-Møller authored 5 years ago and Mads M. Pedersen committed 5 years ago

Updated docker image with texlive-latex-extra for PyWake report

Mikkel Friis-Møller authored 5 years ago and Riccardo Riva committed 5 years ago

Added initial implementation of the predict functions

Fixed pep8

WIP on implementing the evaluation of a surrogate based on a dictionary

Added option for evaluating a surrogate model with a dictionary

Added option for using MinMaxScaler and not only StandardScaler. Improved documentation

Added one more surrogate type, fixed output type and PEP8

Initial implementation of predict_gradient for OpenTURNS metamodels. Improved documentation and clarified names

Bugfix to predict functions and refractoring

Bugfix and improved exceptions

Added documentation for finite differences (to be implemented)

Added implementation for the step size parameters

Added implementation for finite differences

Added step size and output scaling.

Fixed TypeError: 'staticmethod' object is not callable
thanks to https://stackoverflow.com/a/41921291/3676517

Removed class SurrogateEvaluation and placed all of its content in the load module

new structure to allow for load constraints

finished new structure and updated tests

included aggregated_cost.py

Added function to compute the relative and absolute errors

removed try except from load constraints and added setup requirement: openturns vers. >= 1.13

working on ex 8

Added SurrogateModel class

passing input vals to load component to get the shapes right

Added  warning for extrapolation

Added first test for load module

Added test and fixed a bug

Added another test

Added test for SurrogateModel class

Added first test for predict_gradient() and fixed a bug

Added test for PCE

Fixed test_PCE()

Added test for MLPRegressor

Improved test

Added test for wrong input type

Added test for input domain boundary

Added test for unknown model type

Improved test for PCE and neural network

Added test for input and output scaler

Fixed predict_gradient() if there is an input_scaler. Remove code for finite differences, since they will be computed with OpenMDAO

Fixed ouput_scaler in predict_gradient()

added test of load constraint

Added more test

Improved test

coverage up + pep8

added docstrings

pep8

fixed loose connection of post constraint in some cases

Fixed behavior of scikit-learn neural network in case of 1D output

PEP8

Merge branch 'electrical_cost' of https://gitlab.windenergy.dtu.dk/TOPFARM/Topfarm2 into electrical_cost