grayboxes

grayboxes contributes to the creation and evaluation of white, gray, and black box models of physical and chemical transport phenomena.

Gray box models

• are hybrids of theory-driven and data-driven submodels,
• can have adjustable degrees of transparency (the more transparent, the more theory-driven),
• are compatible with all operations of the grayboxes library:
  • Forward simulation
  • Minimization/maximization
  • Inverse problem solution
  • Sensitivity analysis.

grayboxes is the base of the extension package whiteboxes for the implementation of distributed theoretical submodels. whiteboxes-based models are compatible with the model operations of the grayboxes package; see the figure below.

[Link to grayboxes Wiki]

Content

grayboxes
    Training of gray box models, sensitivity analysis
    Optimization and inverse problem solution with white box, gray box and black box models

test
    Module tests

doc
    Figures and manuals used in wiki

       

Installation

git clone https://github.com/dwweiss/grayboxes.git
# ... change to grayboxes-master directory
python3 setup.py install --user

[Link] to the package installation procedure on windows.

Alternatively, all files in the zip file can be copied in the current working directory of the actual application. Press [Clone and Download] and select [Download Zip].

Dependencies

• Modules lightgray and minimum are dependent on package modestga [MGA18]
• Module neuralnlb is dependent on package neurolab [NLB15]
• Module neuraltfl is dependent on package tensorflow [ABA15]
• Module neuraltor is dependent on package torch [ABA15]

As an alternative to installation with setup.py, manual installation of the needed packages can be done with pip:

 pip3 install tensorflow=2.2.2 neurolab matplotlib modestga numpy pandas scipy torch