Learn R Programming

deepregression (version 2.2.0)

Fitting Deep Distributional Regression

Description

Allows for the specification of semi-structured deep distributional regression models which are fitted in a neural network as proposed by Ruegamer et al. (2023) . Predictors can be modeled using structured (penalized) linear effects, structured non-linear effects or using an unstructured deep network model.

Copy Link

Version

Install

install.packages('deepregression')

Monthly Downloads

351

Version

2.2.0

License

GPL-3

Maintainer

David Ruegamer

Last Published

December 2nd, 2024

Functions in deepregression (2.2.0)

Options for formula parsing

family_to_trochd

Character-torch mapping function

Extract attributes/hyper-parameters of the node term

get_weight_by_opname

Function to return weight given model and name

Hadamard-type layers

fitted.drEnsemble

Method for extracting the fitted values of an ensemble

Convenience layer function

loop_through_pfc_and_call_trafo

Function to loop through parsed formulas and apply data trafo

get_partial_effect

Return partial effect of one smooth term

Extract property of gamdata

Extract gam part from wrapped term

get_gamdata_reduced_nr

Extract number in matching table of reduced gam term

layer_dense_module

Function to create custom nn_linear module to overwrite reset_parameters

layer_spline_torch

Function to define spline as Torch layer

from_preds_to_dist

Define Predictor of a Deep Distributional Regression Model

family_to_trafo_torch

Character-to-transformation mapping function

get_help_forward_torch

Helper function to calculate amount of layers Needed when shared layers are used, because of layers have same names

layer_dense_torch

Function to define a torch layer similar to a tf dense layer

Function to return the log_score

get_processor_name

Extract processor name from term

from_distfun_to_dist_torch

Function to define output distribution based on dist_fun

Formula helpers

used by gam_processor

from_preds_to_dist_torch

Define Predictor of a Deep Distributional Regression Model

Function to subset parsed formulas

get_layer_by_opname

Function to return layer given model and name

from_dist_to_loss_torch

Function to transform a distribution layer output into a loss function

get_ensemble_distribution

Obtain the conditional ensemble distribution

Extract term names from the parsed formula content

creates a generator for training

Function to compute adjusted penalty when orthogonalizing

Options for orthogonalization

handle_gam_term

Function to define smoothness and call mgcv's smooth constructor

Generate folds for CV out of one hot encoded matrix

simplyconnected_layer_torch

Hadamard-type layers torch

from_dist_to_loss

Function to transform a distritbution layer output into a loss function

get_distribution

Function to return the fitted distribution

Extract variable from term

Character-tfd mapping function

Extract variables from wrapped node term

Plot CV results from deepregression

Pre-calculate all gam parts from the list of formulas

import_packages

Function to import required packages

Extract terms defined by specials in formula

import_tf_dependings

Function to import required packages for tensorflow @import tensorflow tfprobability keras

Function to define an optimizer combining multiple optimizers

random effect layer

Function to exclude NA values

get_weight_by_name

Function to retrieve the weights of a structured layer

get_layernr_by_opname

Function to return layer number given model and name

get_layernr_trainable

Function to return layer numbers with trainable weights

Compile a Deep Distributional Regression Model

TensorFlow repeat function which is not available for TF 2.0

plot.deepregression

Generic functions for deepregression models

get_luz_dataset

Helper function to create an function that generates R6 instances of class dataset

Generic function to re-intialize model weights

import_torch_dependings

Function to import required packages for torch @import torch torchvision luz

Row-wise tensor product using TensorFlow

Function to prepare data based on parsed formulas

layer_sparse_batch_normalization

Sparse Batch Normalization layer

Function to define spline as TensorFlow layer

nn_init_no_grad_constant_deepreg

custom nn_linear module to overwrite reset_parameters # nn_init_constant works only if value is scalar; so warmstarts for gam does'not work

NODE/ODTs Layer

Returns the parameter names for a given family

layer_sparse_conv_2d

Sparse 2D Convolutional layer

make_generator_from_matrix

Make a DataGenerator from a data.frame or matrix

penalty_control

Options for penalty setup in the pre-processing

Function to initialize a nn_module Forward functions works with a list. The entries of the list are the input of the subnetworks

prepare_data_torch

Function to additionally prepare data for fit process (torch)

orthog_structured_smooths_Z

Orthogonalize structured term by another matrix

orthog_post_fitting

Orthogonalize a Semi-Structured Model Post-hoc

Options for weights of layers

tf_stride_last_dim_tensor

Function to index tensors last dimension

For using mean squared error via TFP

update_miniconda_deepregression

Function to update miniconda and packages

Compile a Deep Distributional Regression Model (Torch)

prepare_torch_distr_mixdistr

Prepares distributions for mixture process

Control function to define the processor for terms in the formula

Function to index tensors columns

tf_split_multiple

Split tensor in multiple parts

layer_generator

Function that creates layer for each processor

subnetwork_init_torch

Initializes a Subnetwork based on the Processed Additive Predictor

subnetwork_init

Initializes a Subnetwork based on the Processed Additive Predictor

Generic quantile function

Function that takes term and create layer name

Generator function for deepregression objects

prepare_input_list_model

Function to prepare input list for fit process, due to different approaches

prepare_newdata

Function to prepare new data based on parsed formulas

predict_gam_handler

Handler for prediction with gam terms

Generic sd function

reinit_weights.deepregression

Method to re-initialize weights of a "deepregression" model

separate_define_relation

Function to define orthogonalization connections in the formula

stop_iter_cv_result

Function to get the stoppting iteration from CV

Implementation of a zero-inflated negbinom distribution for TFP

Implementation of a zero-inflated poisson distribution for TFP

Function to create (custom) family

Function to create mgcv-type penalty

collect_distribution_parameters

Character-to-parameter collection function needed for mixture of same distribution (torch)

layer_add_identity

Convenience layer function

check_input_args_fit

Function to check if inputs are supported by corresponding fit function

choose_kernel_initializer_torch

Function to choose a kernel initializer for a torch layer

Convenience function to extract penalty matrix and value

check_and_install

Function to check python environment and install necessary packages

coef.drEnsemble

Method for extracting ensemble coefficient estimates

combine_penalties

Function to combine two penalties

create_family_torch

Function to create (custom) family

Families for deepregression

Fitting Semi-Structured Deep Distributional Regression

Generic cv function

extract_pure_gam_part

Extract the smooth term from a deepregression term specification

Generic deep ensemble function

distfun_to_dist

Function to define output distribution based on dist_fun

ensemble.deepregression

Ensembling deepregression models

family_to_trafo

Character-to-transformation mapping function