Model

This part of the package contains the stochastic models as well as class for achieving specific goals.

Tasks

The first part is the tasks. At present, there is only one task but it will be interesting to add more of them in the future like quantile regression or binary classification. The aim of the task is to take one stochastic model as attribute and and make all the preprocessing and postprocessing steps needeed for training and testing the models. For instance, GaussianRegression allows the user to scale the input data and unscale the predicted mean and variance.

Here is the list of tasks :

• GaussianRegression

Modules

The second part is the modules. The modules are the building blocks of the stochastic models. Custom layers, activations and models are implemented in order to construct stochastic models. The layers are used to construct the stochastic models and the activations can be used by users to specify their own models dedicated for their specific needs. For more details, an example of utilization of each model is provided in their documentation.

Here is the list of modules :

• Custom Layers
  • Dense2Dto3D
  • Dense3Dto3D
  • Dense3Dto2D
  • Dense3Dto4D
• Custom Activations
  • MeanVarianceActivation
• Custom Models
  • Stochastic Model
  • MVEM
  • DeepEnsemble
  • SWAG
  • MultiSWAG
  • Orthonormal Certificates