rcognita.models

This module contains model classes. These can be used in system dynamics fitting, critic and other tasks

Updates to come.

Functions

force_positive_def(func)

Classes

LookupTable(*dims)
Model()	Blueprint of a model.
ModelBiquadForm(weights)	Bi-quadratic form.
ModelGaussianConditional([…])	Gaussian probability distribution model with weights[0] being an expectation vector and weights[1] being a covariance matrix.
ModelNN()	Class of pytorch neural network models.
ModelQuadForm([weights])	Quadratic form.
ModelQuadLin(dim_input[, weight_min, …])	Quadratic-linear model.
ModelQuadMix(dim_input[, weight_min, weight_max])
ModelQuadNoMix(dim_input[, …])	Quadratic model (no mixed terms).
ModelQuadNoMix2D(dim_input[, …])	Quadratic model (no mixed terms).
ModelQuadNoMixTorch(dim_observation, dim_action)	pytorch neural network of one layer: fully connected.
ModelQuadratic(dim_input[, …])	Quadratic model.
ModelQuadraticSquared(dim_input[, …])	Quadratic model.
ModelSS(A, B, C, D, initial_guessest)
ModelWeightContainer(dim_output[, weights_init])	Trivial model, which is typically used in actor in which actions are being optimized directly.