Help for package survivalmodels

Title:

Models for Survival Analysis

Version:

0.1.191

Description:

Implementations of classical and machine learning models for survival analysis, including deep neural networks via 'keras' and 'tensorflow'. Each model includes a separated fit and predict interface with consistent prediction types for predicting risk or survival probabilities. Models are either implemented from 'Python' via 'reticulate' https://CRAN.R-project.org/package=reticulate, from code in GitHub packages, or novel implementations using 'Rcpp' https://CRAN.R-project.org/package=Rcpp. Neural networks are implemented from the 'Python' package 'pycox' https://github.com/havakv/pycox.

License:

MIT + file LICENSE

URL:

https://github.com/RaphaelS1/survivalmodels/

BugReports:

https://github.com/foucher-y/survivalmodels/issues

Imports:

Rcpp (≥ 1.0.5)

Suggests:

keras (≥ 2.11.0), pseudo, reticulate, survival

LinkingTo:

Rcpp

Encoding:

UTF-8

NeedsCompilation:

yes

Packaged:

2024-03-18 21:10:19 UTC; foucher-y

Author:

Raphael Sonabend

[aut], Yohann Foucher

[cre]

Maintainer:

Yohann Foucher <yohann.foucher@univ-poitiers.fr>

Repository:

CRAN

Date/Publication:

2024-03-19 16:50:03 UTC

survivalmodels: Models for Survival Analysis

Description

survivalmodels implements classical and machine learning models for survival analysis that either do not already exist in R or for more efficient implementations.

Author(s)

Maintainer: Yohann Foucher yohann.foucher@univ-poitiers.fr (ORCID) Authors:

Raphael Sonabend (ORCID)

Build a Keras Multilayer Perceptron

Description

Utility function to build a Keras MLP.

Usage

build_keras_net(
  n_in,
  n_out,
  nodes = c(32L, 32L),
  layer_pars = list(),
  activation = "linear",
  act_pars = list(),
  dropout = 0.1,
  batch_norm = TRUE,
  batch_pars = list()
)

Arguments

n_in

(integer(1))
Number of input features.

n_out

(integer(1))
Number of targets.

nodes

(numeric())
Hidden nodes in network, each element in vector represents number of hidden nodes in respective layer.

layer_pars

(list())
Arguments passed to keras::layer_dense.

activation

(character(1))
Activation function passed to keras::layer_activation. Default is linear.

act_pars

(list())
Parameters for activation function, see keras::layer_activation.

dropout

(numeric(1))
Optional dropout layer, if NULL then no dropout layer added otherwise either same dropout will be added to all layers.

batch_norm

(logical(1))
If TRUE (default) then batch normalisation is applied to all layers.

batch_pars

(list())
Parameters for batch normalisation, see keras::layer_batch_normalization.

Details

This function is a helper for R users with less Python experience. Currently it is limited to simple MLPs and with identical layers. More advanced networks will require manual creation with keras.

Value

No return value.

Build a Pytorch Multilayer Perceptron

Description

Utility function to build an MLP with a choice of activation function and weight initialization with optional dropout and batch normalization.

Usage

build_pytorch_net(
  n_in,
  n_out,
  nodes = c(32, 32),
  activation = "relu",
  act_pars = list(),
  dropout = 0.1,
  bias = TRUE,
  batch_norm = TRUE,
  batch_pars = list(eps = 1e-05, momentum = 0.1, affine = TRUE),
  init = "uniform",
  init_pars = list()
)

Arguments

n_in

(integer(1))
Number of input features.

n_out

(integer(1))
Number of targets.

nodes

(numeric())
Hidden nodes in network, each element in vector represents number of hidden nodes in respective layer.

activation

(character(1)|list())
Activation function, can either be a single character and the same function is used in all layers, or a list of length length(nodes). See get_pycox_activation for options.

act_pars

(list())
Passed to get_pycox_activation.

dropout

(numeric())
Optional dropout layer, if NULL then no dropout layer added otherwise either a single numeric which will be added to all layers or a vector of differing drop-out amounts.

bias

(logical(1))
If TRUE (default) then a bias parameter is added to all linear layers.

batch_norm

(logical(1))
If TRUE (default) then batch normalisation is applied to all layers.

batch_pars

(list())
Parameters for batch normalisation, see reticulate::py_help(torch$nn$BatchNorm1d).

init

(character(1))
Weight initialization method. See get_pycox_init for options.

init_pars

(list())
Passed to get_pycox_init.

Details

This function is a helper for R users with less Python experience. Currently it is limited to simple MLPs. More advanced networks will require manual creation with reticulate.

Value

No return value.

Compute Concordance of survivalmodel Risk

Description

A thin wrapper around survival::concordance which essentially just sets reverse = TRUE.

Usage

cindex(risk, truth, ...)

Arguments

risk

(numeric())
Vector of risk predictions from a survivalmodel model (so high risk implies low survival time prediction).

truth

(numeric())
Vector of true survival times, must be same length as risk.

...

(ANY)
Further parameters passed to survival::concordance.

Value

The numeric value of the index.

Examples

if (!requireNamespace("survival", quietly = TRUE)) {
  set.seed(10)
  data <- simsurvdata(20)
  fit <- deepsurv(data = data[1:10, ])
  p <- predict(fit, type = "risk", newdata = data[11:20, ])
  concordance(risk = p, truth = data[11:20, "time"])
}

Cox-Time Survival Neural Network

Description

Cox-Time fits a neural network based on the Cox PH with possibly time-dependent effects.

Usage

coxtime(
  formula = NULL,
  data = NULL,
  reverse = FALSE,
  time_variable = "time",
  status_variable = "status",
  x = NULL,
  y = NULL,
  frac = 0,
  standardize_time = FALSE,
  log_duration = FALSE,
  with_mean = TRUE,
  with_std = TRUE,
  activation = "relu",
  num_nodes = c(32L, 32L),
  batch_norm = TRUE,
  dropout = NULL,
  device = NULL,
  shrink = 0,
  early_stopping = FALSE,
  best_weights = FALSE,
  min_delta = 0,
  patience = 10L,
  batch_size = 256L,
  epochs = 1L,
  verbose = FALSE,
  num_workers = 0L,
  shuffle = TRUE,
  ...
)

Arguments

formula

(formula(1))
Object specifying the model fit, left-hand-side of formula should describe a survival::Surv() object.

data

(data.frame(1))
Training data of data.frame like object, internally is coerced with stats::model.matrix().

reverse

(logical(1))
If TRUE fits estimator on censoring distribution, otherwise (default) survival distribution.

time_variable

(character(1))
Alternative method to call the function. Name of the 'time' variable, required if formula. or x and Y not given.

status_variable

(character(1))
Alternative method to call the function. Name of the 'status' variable, required if formula or x and Y not given.

x

(data.frame(1))
Alternative method to call the function. Required if ⁠formula, time_variable⁠ and status_variable not given. Data frame like object of features which is internally coerced with model.matrix.

y

⁠([survival::Surv()])⁠
Alternative method to call the function. Required if ⁠formula, time_variable⁠ and status_variable not given. Survival outcome of right-censored observations.

frac

(numeric(1))
Fraction of data to use for validation dataset, default is 0 and therefore no separate validation dataset.

standardize_time

(logical(1))
If TRUE, the time outcome is standardized.

log_duration

(logical(1))
If TRUE and standardize_time is TRUE then time variable is log transformed.

with_mean

(logical(1))
If TRUE (default) and standardize_time is TRUE then time variable is centered.

with_std

(logical(1))
If TRUE (default) and standardize_time is TRUE then time variable is scaled to unit variance.

activation

(character(1))
See get_pycox_activation.

num_nodes, batch_norm, dropout

(integer()/logical(1)/numeric(1))
See build_pytorch_net.

device

(integer(1)|character(1))
Passed to pycox.models.Coxtime, specifies device to compute models on.

shrink

(numeric(1))
Passed to pycox.models.Coxtime, shrinkage parameter for regularization.

early_stopping, best_weights, min_delta, patience

⁠(logical(1)/logical(1)/numeric(1)/integer(1)⁠
See get_pycox_callbacks.

batch_size

(integer(1))
Passed to pycox.models.Coxtime.fit, elements in each batch.

epochs

(integer(1))
Passed to pycox.models.Coxtime.fit, number of epochs.

verbose

(logical(1))
Passed to pycox.models.Coxtime.fit, should information be displayed during fitting.

num_workers

(integer(1))
Passed to pycox.models.Coxtime.fit, number of workers used in the dataloader.

shuffle

(logical(1))
Passed to pycox.models.Coxtime.fit, should order of dataset be shuffled?

...

ANY
Passed to get_pycox_optim.

Details

Implemented from the pycox Python package via reticulate. Calls pycox.models.Coxtime.

Value

An object inheriting from class coxtime.

An object of class survivalmodel.

References

Kvamme, H., Borgan, Ø., & Scheel, I. (2019). Time-to-event prediction with neural networks and Cox regression. Journal of Machine Learning Research, 20(129), 1–30.

DeepHit Survival Neural Network

Description

DeepHit fits a neural network based on the PMF of a discrete Cox model. This is the single (non-competing) event implementation.

Usage

deephit(
  formula = NULL,
  data = NULL,
  reverse = FALSE,
  time_variable = "time",
  status_variable = "status",
  x = NULL,
  y = NULL,
  frac = 0,
  cuts = 10,
  cutpoints = NULL,
  scheme = c("equidistant", "quantiles"),
  cut_min = 0,
  activation = "relu",
  custom_net = NULL,
  num_nodes = c(32L, 32L),
  batch_norm = TRUE,
  dropout = NULL,
  device = NULL,
  mod_alpha = 0.2,
  sigma = 0.1,
  early_stopping = FALSE,
  best_weights = FALSE,
  min_delta = 0,
  patience = 10L,
  batch_size = 256L,
  epochs = 1L,
  verbose = FALSE,
  num_workers = 0L,
  shuffle = TRUE,
  ...
)

Arguments

formula

(formula(1))
Object specifying the model fit, left-hand-side of formula should describe a survival::Surv() object.

data

(data.frame(1))
Training data of data.frame like object, internally is coerced with stats::model.matrix().

reverse

(logical(1))
If TRUE fits estimator on censoring distribution, otherwise (default) survival distribution.

time_variable

(character(1))
Alternative method to call the function. Name of the 'time' variable, required if formula. or x and Y not given.

status_variable

(character(1))
Alternative method to call the function. Name of the 'status' variable, required if formula or x and Y not given.

x

y

⁠([survival::Surv()])⁠
Alternative method to call the function. Required if ⁠formula, time_variable⁠ and status_variable not given. Survival outcome of right-censored observations.

frac

(numeric(1))
Fraction of data to use for validation dataset, default is 0 and therefore no separate validation dataset.

cuts

(integer(1))
If discretise is TRUE then determines number of cut-points for discretisation.

cutpoints

(numeric())
Alternative to cuts if discretise is true, provide exact cutpoints for discretisation. cuts is ignored if cutpoints is non-NULL.

scheme

(character(1))
Method of discretisation, either "equidistant" (default) or "quantiles". See reticulate::py_help(pycox$models$LogisticHazard$label_transform) for more detail.

cut_min

(integer(1))
Starting duration for discretisation, see reticulate::py_help(pycox$models$LogisticHazard$label_transform) for more detail.

activation

(character(1))
See get_pycox_activation.

custom_net

(torch.nn.modules.module.Module(1))
Optional custom network built with build_pytorch_net, otherwise default architecture used. Note that if building a custom network the number of output channels depends on cuts or cutpoints.

num_nodes, batch_norm, dropout

(integer()/logical(1)/numeric(1))
See build_pytorch_net.

device

(integer(1)|character(1))
Passed to pycox.models.DeepHitSingle, specifies device to compute models on.

mod_alpha

(numeric(1))
Weighting in (0,1) for combining likelihood (L1) and rank loss (L2). See reference and py_help(pycox$models$DeepHitSingle) for more detail.

sigma

(numeric(1))
From eta in rank loss (L2) of ref. See reference and py_help(pycox$models$DeepHitSingle) for more detail.

early_stopping, best_weights, min_delta, patience

⁠(logical(1)/logical(1)/numeric(1)/integer(1)⁠
See get_pycox_callbacks.

batch_size

(integer(1))
Passed to pycox.models.DeepHitSingle.fit, elements in each batch.

epochs

(integer(1))
Passed to pycox.models.DeepHitSingle.fit, number of epochs.

verbose

(logical(1))
Passed to pycox.models.DeepHitSingle.fit, should information be displayed during fitting.

num_workers

(integer(1))
Passed to pycox.models.DeepHitSingle.fit, number of workers used in the dataloader.

shuffle

(logical(1))
Passed to pycox.models.DeepHitSingle.fit, should order of dataset be shuffled?

...

ANY
Passed to get_pycox_optim.

Details

Implemented from the pycox Python package via reticulate. Calls pycox.models.DeepHitSingle.

Value

An object inheriting from class deephit.

An object of class survivalmodel.

References

Changhee Lee, William R Zame, Jinsung Yoon, and Mihaela van der Schaar. Deephit: A deep learning approach to survival analysis with competing risks. In Thirty-Second AAAI Conference on Artificial Intelligence, 2018. http://medianetlab.ee.ucla.edu/papers/AAAI_2018_DeepHit

DeepSurv Survival Neural Network

Description

DeepSurv neural fits a neural network based on the partial likelihood from a Cox PH.

Usage

deepsurv(
  formula = NULL,
  data = NULL,
  reverse = FALSE,
  time_variable = "time",
  status_variable = "status",
  x = NULL,
  y = NULL,
  frac = 0,
  activation = "relu",
  num_nodes = c(32L, 32L),
  batch_norm = TRUE,
  dropout = NULL,
  device = NULL,
  early_stopping = FALSE,
  best_weights = FALSE,
  min_delta = 0,
  patience = 10L,
  batch_size = 256L,
  epochs = 1L,
  verbose = FALSE,
  num_workers = 0L,
  shuffle = TRUE,
  ...
)

Arguments

formula

(formula(1))
Object specifying the model fit, left-hand-side of formula should describe a survival::Surv() object.

data

(data.frame(1))
Training data of data.frame like object, internally is coerced with stats::model.matrix().

reverse

(logical(1))
If TRUE fits estimator on censoring distribution, otherwise (default) survival distribution.

time_variable

(character(1))
Alternative method to call the function. Name of the 'time' variable, required if formula. or x and Y not given.

status_variable

(character(1))
Alternative method to call the function. Name of the 'status' variable, required if formula or x and Y not given.

x

y

⁠([survival::Surv()])⁠
Alternative method to call the function. Required if ⁠formula, time_variable⁠ and status_variable not given. Survival outcome of right-censored observations.

frac

(numeric(1))
Fraction of data to use for validation dataset, default is 0 and therefore no separate validation dataset.

activation

(character(1))
See get_pycox_activation.

num_nodes, batch_norm, dropout

(integer()/logical(1)/numeric(1))
See build_pytorch_net.

device

(integer(1)|character(1))
Passed to pycox.models.CoxPH, specifies device to compute models on.

early_stopping, best_weights, min_delta, patience

⁠(logical(1)/logical(1)/numeric(1)/integer(1)⁠
See get_pycox_callbacks.

batch_size

(integer(1))
Passed to pycox.models.CoxPH.fit, elements in each batch.

epochs

(integer(1))
Passed to pycox.models.CoxPH.fit, number of epochs.

verbose

(logical(1))
Passed to pycox.models.CoxPH.fit, should information be displayed during fitting.

num_workers

(integer(1))
Passed to pycox.models.CoxPH.fit, number of workers used in the dataloader.

shuffle

(logical(1))
Passed to pycox.models.CoxPH.fit, should order of dataset be shuffled?

...

ANY
Passed to get_pycox_optim.

Details

Implemented from the pycox Python package via reticulate. Calls pycox.models.CoxPH.

Value

An object inheriting from class deepsurv.

An object of class survivalmodel.

References

Katzman, J. L., Shaham, U., Cloninger, A., Bates, J., Jiang, T., & Kluger, Y. (2018). DeepSurv: personalized treatment recommender system using a Cox proportional hazards deep neural network. BMC Medical Research Methodology, 18(1), 24. https://doi.org/10.1186/s12874-018-0482-1

Get Keras Optimizer

Description

Utility function to construct optimiser from keras, primarily for internal use.

Usage

get_keras_optimizer(
  optimizer = "adam",
  lr = 0.001,
  beta_1 = 0.9,
  beta_2 = 0.999,
  epsilon = 1e-07,
  decay = NULL,
  clipnorm = NULL,
  clipvalue = NULL,
  momentum = 0,
  nesterov = FALSE,
  rho = 0.95,
  global_clipnorm = NULL,
  use_ema = FALSE,
  ema_momentum = 0.99,
  ema_overwrite_frequency = NULL,
  jit_compile = TRUE,
  initial_accumultator_value = 0.1,
  amsgrad = FALSE,
  lr_power = -0.5,
  l1_regularization_strength = 0,
  l2_regularization_strength = 0,
  l2_shrinkage_regularization_strength = 0,
  beta = 0,
  centered = FALSE
)

Arguments

optimizer

(character(1))
Optimizer to construct, see details for those available. Default is "adam".

lr

(numeric(1))
Learning rate passed to all optimizers.

beta_1, beta_2

(numeric(1))
Passed to adamax, adam, and nadam.

epsilon

(numeric(1))
Passed to adadelta, adagrad, adam, adamax, nadam, rmsprop

decay, clipnorm, clipvalue, global_clipnorm

(numeric(1))
Passed to all optimizers.

momentum

(numeric(1))
Passed to rmsprop and sgd.

nesterov

(logical(1))
Passed to sgd.

rho

(numeric(1))
Passed to adadelta and rmsprop.

use_ema, jit_compile

(logical(1))
Passed to all optimizers.

ema_momentum, ema_overwrite_frequency

(numeric(1))
Passed to all optimizers.

initial_accumultator_value

(numeric(1))
Passed to adagrad and ftrl.

amsgrad

(logical(1))
Passed to adam and sgd.

lr_power, l1_regularization_strength, l2_regularization_strength, l2_shrinkage_regularization_strength, beta

(numeric(1))
Passed to ftrl.

centered

(logical(1))
Passed to rmsprop.

Details

Implemented optimizers are

Value

No return value.

Get Pytorch Activation Function

Description

Helper function to return a class or constructed object for pytorch activation function from torch.nn.modules.activation.

Usage

get_pycox_activation(
  activation = "relu",
  construct = TRUE,
  alpha = 1,
  dim = NULL,
  lambd = 0.5,
  min_val = -1,
  max_val = 1,
  negative_slope = 0.01,
  num_parameters = 1L,
  init = 0.25,
  lower = 1/8,
  upper = 1/3,
  beta = 1,
  threshold = 20,
  value = 20
)

Arguments

activation

(character(1))
Activation function method, see details for list of implemented methods.

construct

(logical(1))
If TRUE (default) returns constructed object, otherwise a class.

alpha

(numeric(1))
Passed to celu and elu.

dim

(integer(1))
Passed to glu, logsoftmax, softmax, and

lambd

(numeric(1))
Passed to hardshrink and softshrink.

min_val, max_val

(numeric(1))
Passed to hardtanh.

negative_slope

(numeric(1))
Passed to leakyrelu.

num_parameters

(integer(1))
Passed to prelu.

init

(numeric(1))
Passed to prelu.

lower, upper

(numeric(1))
Passed to rrelu.

beta

(numeric(1))
Passed to softplus.

threshold

(numeric(1))
Passed to softplus and threshold.

value

(numeric(1))
Passed to threshold.

Details

Implemented methods (with help pages) are

"celu"
reticulate::py_help(torch$nn$modules$activation$CELU)
"elu"
reticulate::py_help(torch$nn$modules$activation$ELU)
"gelu"
reticulate::py_help(torch$nn$modules$activation$GELU)
"glu"
reticulate::py_help(torch$nn$modules$activation$GLU)
"hardshrink"
reticulate::py_help(torch$nn$modules$activation$Hardshrink)
"hardsigmoid"
reticulate::py_help(torch$nn$modules$activation$Hardsigmoid)
"hardswish"
reticulate::py_help(torch$nn$modules$activation$Hardswish)
"hardtanh"
reticulate::py_help(torch$nn$modules$activation$Hardtanh)
"relu6"
reticulate::py_help(torch$nn$modules$activation$ReLU6)
"leakyrelu"
reticulate::py_help(torch$nn$modules$activation$LeakyReLU)
"logsigmoid"
reticulate::py_help(torch$nn$modules$activation$LogSigmoid)
"logsoftmax"
reticulate::py_help(torch$nn$modules$activation$LogSoftmax)
"prelu"
reticulate::py_help(torch$nn$modules$activation$PReLU)
"rrelu"
reticulate::py_help(torch$nn$modules$activation$RReLU)
"relu"
reticulate::py_help(torch$nn$modules$activation$ReLU)
"selu"
reticulate::py_help(torch$nn$modules$activation$SELU)
"sigmoid"
reticulate::py_help(torch$nn$modules$activation$Sigmoid)
"softmax"
reticulate::py_help(torch$nn$modules$activation$Softmax)
"softmax2d"
reticulate::py_help(torch$nn$modules$activation$Softmax2d)
"softmin"
reticulate::py_help(torch$nn$modules$activation$Softmin)
"softplus"
reticulate::py_help(torch$nn$modules$activation$Softplus)
"softshrink"
reticulate::py_help(torch$nn$modules$activation$Softshrink)
"softsign"
reticulate::py_help(torch$nn$modules$activation$Softsign)
"tanh"
reticulate::py_help(torch$nn$modules$activation$Tanh)
"tanhshrink"
reticulate::py_help(torch$nn$modules$activation$Tanhshrink)
"threshold"
reticulate::py_help(torch$nn$modules$activation$Threshold)

Value

No return value.

Get Torchtuples Callbacks

Description

Helper function to return torchtuples callbacks from torchtuples.callbacks.

Usage

get_pycox_callbacks(
  early_stopping = FALSE,
  best_weights = FALSE,
  min_delta = 0,
  patience = 10L
)

Arguments

early_stopping

(logical(1))
If TRUE then constructs ⁠torchtuples.callbacks,EarlyStopping⁠.

best_weights

(logical(1))
If TRUE then returns torchtuples.callbacks.BestWeights. Ignored if early_stopping is TRUE.

min_delta

(numeric(1))
Passed to torchtuples.callbacks.EarlyStopping.

patience

(integer(1))
Passed to torchtuples.callbacks.EarlyStopping.

Value

No return value.

Get Pytorch Weight Initialization Method

Description

Helper function to return a character string with a populated pytorch weight initializer method from torch.nn.init. Used in build_pytorch_net to define a weighting function.

Usage

get_pycox_init(
  init = "uniform",
  a = 0,
  b = 1,
  mean = 0,
  std = 1,
  val,
  gain = 1,
  mode = c("fan_in", "fan_out"),
  non_linearity = c("leaky_relu", "relu")
)

Arguments

init

(character(1))
Initialization method, see details for list of implemented methods.

a

(numeric(1))
Passed to uniform, kaiming_uniform, and kaiming_normal.

b

(numeric(1))
Passed to uniform.

mean, std

(numeric(1))
Passed to normal.

val

(numeric(1))
Passed to constant.

gain

(numeric(1))
Passed to xavier_uniform, xavier_normal, and orthogonal.

mode

(character(1))
Passed to kaiming_uniform and kaiming_normal, one of fan_in (default) and fan_out.

non_linearity

(character(1))
Passed to kaiming_uniform and kaiming_normal, one of leaky_relu (default) and relu.

Details

Implemented methods (with help pages) are

"uniform"
reticulate::py_help(torch$nn$init$uniform_)
"normal"
reticulate::py_help(torch$nn$init$normal_)
"constant"
reticulate::py_help(torch$nn$init$constant_)
"xavier_uniform"
reticulate::py_help(torch$nn$init$xavier_uniform_)
"xavier_normal"
reticulate::py_help(torch$nn$init$xavier_normal_)
"kaiming_uniform"
reticulate::py_help(torch$nn$init$kaiming_uniform_)
"kaiming_normal"
reticulate::py_help(torch$nn$init$kaiming_normal_)
"orthogonal"
reticulate::py_help(torch$nn$init$orthogonal_)

Value

No return value.

Get Pytorch Optimizer

Description

Helper function to return a constructed pytorch optimizer from torch.optim.

Usage

get_pycox_optim(
  optimizer = "adam",
  net,
  rho = 0.9,
  eps = 1e-08,
  lr = 1,
  weight_decay = 0,
  learning_rate = 0.01,
  lr_decay = 0,
  betas = c(0.9, 0.999),
  amsgrad = FALSE,
  lambd = 1e-04,
  alpha = 0.75,
  t0 = 1e+06,
  momentum = 0,
  centered = TRUE,
  etas = c(0.5, 1.2),
  step_sizes = c(1e-06, 50),
  dampening = 0,
  nesterov = FALSE
)

Arguments

optimizer

(character(1))
Optimizer, see details for list of implemented methods.

net

(torch.nn.modules.module.Module)
Network architecture, can be built from build_pytorch_net.

rho, lr, lr_decay

(numeric(1))
Passed to adadelta.

eps

(numeric(1))
Passed to all methods except asgd, rprop, and sgd.

weight_decay

(numeric(1))
Passed to all methods except rprop and sparse_adam.

learning_rate

(numeric(1))
Passed to all methods except adadelta.

betas

(numeric(2))
Passed to adam, adamax, adamw, and sparse_adam.

amsgrad

(logical(1))
Passed to adam and adamw.

lambd, t0

(numeric(1))
Passed to asgd.

alpha

(numeric(1))
Passed to asgd and rmsprop.

momentum

(numeric(1))
Passed to rmsprop and sgd.

centered

(logical(1))
Passed to rmsprop.

etas, step_sizes

(numeric(2))
Passed to rprop.

dampening

(numeric(1))
Passed to sgd.

nesterov

(logical(1))
Passed to sgd.

Details

Implemented methods (with help pages) are

"adadelta"
reticulate::py_help(torch$optim$Adadelta)
"adagrad"
reticulate::py_help(torch$optim$Adagrad)
"adam"
reticulate::py_help(torch$optim$Adam)
"adamax"
reticulate::py_help(torch$optim$Adamax)
"adamw"
reticulate::py_help(torch$optim$AdamW)
"asgd"
reticulate::py_help(torch$optim$ASGD)
"rmsprop"
reticulate::py_help(torch$optim$RMSprop)
"rprop"
reticulate::py_help(torch$optim$Rprop)
"sgd"
reticulate::py_help(torch$optim$SGD)
"sparse_adam"
reticulate::py_help(torch$optim$SparseAdam)

Value

No return value.

Install Keras and Tensorflow

Description

Stripped back version of keras::install_keras. Note the default for pip is changed to TRUE.

Usage

install_keras(
  method = "auto",
  conda = "auto",
  pip = TRUE,
  install_tensorflow = FALSE,
  ...
)

Arguments

method, conda, pip

See reticulate::py_install.

install_tensorflow

If TRUE installs the dependency tensorflow package as well.

...

Passed to reticulate::py_install.

Value

No return value.

Install Pycox With Reticulate

Description

Installs the python 'pycox' package via reticulate. Note the default for pip is changed to TRUE.

Usage

install_pycox(
  method = "auto",
  conda = "auto",
  pip = TRUE,
  install_torch = FALSE,
  ...
)

Arguments

method, conda, pip

See reticulate::py_install.

install_torch

If TRUE installs the dependency torch package as well.

...

Passed to reticulate::py_install.

Value

No return value.

Install Torch With Reticulate

Description

Installs the python 'torch' package via reticulate. Note the default for pip is changed to TRUE.

Usage

install_torch(method = "auto", conda = "auto", pip = TRUE)

Arguments

method, conda, pip

See reticulate::py_install

Value

No return value.

Logistic-Hazard Survival Neural Network

Description

Logistic-Hazard fits a discrete neural network based on a cross-entropy loss and predictions of a discrete hazard function, also known as Nnet-Survival.

Usage

loghaz(
  formula = NULL,
  data = NULL,
  reverse = FALSE,
  time_variable = "time",
  status_variable = "status",
  x = NULL,
  y = NULL,
  frac = 0,
  cuts = 10,
  cutpoints = NULL,
  scheme = c("equidistant", "quantiles"),
  cut_min = 0,
  activation = "relu",
  custom_net = NULL,
  num_nodes = c(32L, 32L),
  batch_norm = TRUE,
  dropout = NULL,
  device = NULL,
  early_stopping = FALSE,
  best_weights = FALSE,
  min_delta = 0,
  patience = 10L,
  batch_size = 256L,
  epochs = 1L,
  verbose = FALSE,
  num_workers = 0L,
  shuffle = TRUE,
  ...
)

Arguments

formula

(formula(1))
Object specifying the model fit, left-hand-side of formula should describe a survival::Surv() object.

data

(data.frame(1))
Training data of data.frame like object, internally is coerced with stats::model.matrix().

reverse

(logical(1))
If TRUE fits estimator on censoring distribution, otherwise (default) survival distribution.

time_variable

(character(1))
Alternative method to call the function. Name of the 'time' variable, required if formula. or x and Y not given.

status_variable

(character(1))
Alternative method to call the function. Name of the 'status' variable, required if formula or x and Y not given.

x

y

⁠([survival::Surv()])⁠
Alternative method to call the function. Required if ⁠formula, time_variable⁠ and status_variable not given. Survival outcome of right-censored observations.

frac

(numeric(1))
Fraction of data to use for validation dataset, default is 0 and therefore no separate validation dataset.

cuts

(integer(1))
If discretise is TRUE then determines number of cut-points for discretisation.

cutpoints

(numeric())
Alternative to cuts if discretise is true, provide exact cutpoints for discretisation. cuts is ignored if cutpoints is non-NULL.

scheme

(character(1))
Method of discretisation, either "equidistant" (default) or "quantiles". See reticulate::py_help(pycox$models$LogisticHazard$label_transform) for more detail.

cut_min

(integer(1))
Starting duration for discretisation, see reticulate::py_help(pycox$models$LogisticHazard$label_transform) for more detail.

activation

(character(1))
See get_pycox_activation.

custom_net

num_nodes, batch_norm, dropout

(integer()/logical(1)/numeric(1))
See build_pytorch_net.

device

(integer(1)|character(1))
Passed to pycox.models.LogisticHazard, specifies device to compute models on.

early_stopping, best_weights, min_delta, patience

⁠(logical(1)/logical(1)/numeric(1)/integer(1)⁠
See get_pycox_callbacks.

batch_size

(integer(1))
Passed to pycox.models.LogisticHazard.fit, elements in each batch.

epochs

(integer(1))
Passed to pycox.models.LogisticHazard.fit, number of epochs.

verbose

(logical(1))
Passed to pycox.models.LogisticHazard.fit, should information be displayed during fitting.

num_workers

(integer(1))
Passed to pycox.models.LogisticHazard.fit, number of workers used in the dataloader.

shuffle

(logical(1))
Passed to pycox.models.LogisticHazard.fit, should order of dataset be shuffled?

...

ANY
Passed to get_pycox_optim.

Details

Implemented from the pycox Python package via reticulate. Calls pycox.models.LogisticHazard.

Value

An object inheriting from class loghaz.

An object of class survivalmodel.

References

Gensheimer, M. F., & Narasimhan, B. (2018). A Simple Discrete-Time Survival Model for Neural Networks, 1–17. https://doi.org/arXiv:1805.00917v3

Kvamme, H., & Borgan, Ø. (2019). Continuous and discrete-time survival prediction with neural networks. https://doi.org/arXiv:1910.06724.

PC-Hazard Survival Neural Network

Description

Logistic-Hazard fits a discrete neural network based on a cross-entropy loss and predictions of a discrete hazard function, also known as Nnet-Survival.

Usage

pchazard(
  formula = NULL,
  data = NULL,
  reverse = FALSE,
  time_variable = "time",
  status_variable = "status",
  x = NULL,
  y = NULL,
  frac = 0,
  cuts = 10,
  cutpoints = NULL,
  scheme = c("equidistant", "quantiles"),
  cut_min = 0,
  activation = "relu",
  custom_net = NULL,
  num_nodes = c(32L, 32L),
  batch_norm = TRUE,
  reduction = c("mean", "none", "sum"),
  dropout = NULL,
  device = NULL,
  early_stopping = FALSE,
  best_weights = FALSE,
  min_delta = 0,
  patience = 10L,
  batch_size = 256L,
  epochs = 1L,
  verbose = FALSE,
  num_workers = 0L,
  shuffle = TRUE,
  ...
)

Arguments

formula

(formula(1))
Object specifying the model fit, left-hand-side of formula should describe a survival::Surv() object.

data

(data.frame(1))
Training data of data.frame like object, internally is coerced with stats::model.matrix().

reverse

(logical(1))
If TRUE fits estimator on censoring distribution, otherwise (default) survival distribution.

time_variable

(character(1))
Alternative method to call the function. Name of the 'time' variable, required if formula. or x and Y not given.

status_variable

(character(1))
Alternative method to call the function. Name of the 'status' variable, required if formula or x and Y not given.

x

y

⁠([survival::Surv()])⁠
Alternative method to call the function. Required if ⁠formula, time_variable⁠ and status_variable not given. Survival outcome of right-censored observations.

frac

(numeric(1))
Fraction of data to use for validation dataset, default is 0 and therefore no separate validation dataset.

cuts

(integer(1))
If discretise is TRUE then determines number of cut-points for discretisation.

cutpoints

(numeric())
Alternative to cuts if discretise is true, provide exact cutpoints for discretisation. cuts is ignored if cutpoints is non-NULL.

scheme

(character(1))
Method of discretisation, either "equidistant" (default) or "quantiles". See reticulate::py_help(pycox$models$LogisticHazard$label_transform) for more detail.

cut_min

(integer(1))
Starting duration for discretisation, see reticulate::py_help(pycox$models$LogisticHazard$label_transform) for more detail.

activation

(character(1))
See get_pycox_activation.

custom_net

num_nodes, batch_norm, dropout

(integer()/logical(1)/numeric(1))
See build_pytorch_net.

reduction

(character(1))
How to reduce the loss, see to reticulate::py_help(pycox$models$loss$NLLPCHazardLoss).

device

(integer(1)|character(1))
Passed to pycox.models.PCHazard, specifies device to compute models on.

early_stopping, best_weights, min_delta, patience

⁠(logical(1)/logical(1)/numeric(1)/integer(1)⁠
See get_pycox_callbacks.

batch_size

(integer(1))
Passed to pycox.models.PCHazard.fit, elements in each batch.

epochs

(integer(1))
Passed to pycox.models.PCHazard.fit, number of epochs.

verbose

(logical(1))
Passed to pycox.models.PCHazard.fit, should information be displayed during fitting.

num_workers

(integer(1))
Passed to pycox.models.PCHazard.fit, number of workers used in the dataloader.

shuffle

(logical(1))
Passed to pycox.models.PCHazard.fit, should order of dataset be shuffled?

...

ANY
Passed to get_pycox_optim.

Details

Implemented from the pycox Python package via reticulate. Calls pycox.models.PCHazard.

Value

An object inheriting from class pchazard.

An object of class survivalmodel.

References

Kvamme, H., & Borgan, Ø. (2019). Continuous and discrete-time survival prediction with neural networks. https://doi.org/arXiv:1910.06724.

Predict Method for pycox Neural Networks

Description

Predicted values from a fitted pycox ANN.

Usage

## S3 method for class 'pycox'
predict(
  object,
  newdata,
  batch_size = 256L,
  num_workers = 0L,
  interpolate = FALSE,
  inter_scheme = c("const_hazard", "const_pdf"),
  sub = 10L,
  type = c("survival", "risk", "all"),
  ...
)

Arguments

object

(pycox(1))
Object of class inheriting from "pycox".

newdata

(data.frame(1))
Testing data of data.frame like object, internally is coerced with stats::model.matrix(). If missing then training data from fitted object is used.

batch_size

(integer(1))
Passed to pycox.models.X.fit, elements in each batch.

num_workers

(integer(1))
Passed to pycox.models.X.fit, number of workers used in the dataloader.

interpolate

(logical(1))
For models deephit and loghaz, should predictions be linearly interpolated? Ignored for other models.

inter_scheme

(character(1))
If interpolate is TRUE then the scheme for interpolation, see reticulate::py_help(py_help(pycox$models$DeepHitSingle$interpolate)) for further details.

sub

(integer(1))
If interpolate is TRUE or model is loghaz, number of sub-divisions for interpolation. See reticulate::py_help(py_help(pycox$models$DeepHitSingle$interpolate))' for further details.

type

(character(1))
Type of predicted value. Choices are survival probabilities over all time-points in training data ("survival") or a relative risk ranking ("risk"), which is the negative mean survival time so higher rank implies higher risk of event, or both ("all").

...

ANY
Currently ignored.

Value

A numeric if type = "risk", a matrix if type = "survival" where entries are survival probabilities with rows of observations and columns are time-points.

Prepare Data for Pycox Model Training

Description

Utility function to prepare data for training in a Pycox model. Generally used internally only.

Usage

pycox_prepare_train_data(
  x_train,
  y_train,
  frac = 0,
  standardize_time = FALSE,
  log_duration = FALSE,
  with_mean = TRUE,
  with_std = TRUE,
  discretise = FALSE,
  cuts = 10L,
  cutpoints = NULL,
  scheme = c("equidistant", "quantiles"),
  cut_min = 0L,
  model = c("coxtime", "deepsurv", "deephit", "loghaz", "pchazard")
)

Arguments

x_train

(matrix(1))
Training covariates.

y_train

(matrix(1))
Training outcomes.

frac

(numeric(1))
Fraction of data to use for validation dataset, default is 0 and therefore no separate validation dataset.

standardize_time

(logical(1))
If TRUE, the time outcome to be standardized. For use with coxtime.

log_duration

(logical(1))
If TRUE and standardize_time is TRUE then time variable is log transformed.

with_mean

(logical(1))
If TRUE (default) and standardize_time is TRUE then time variable is centered.

with_std

(logical(1))
If TRUE (default) and standardize_time is TRUE then time variable is scaled to unit variance.

discretise

(logical(1))
If TRUE then time is discretised. For use with the models deephit, pchazard, and loghaz.

cuts

(integer(1))
If discretise is TRUE then determines number of cut-points for discretisation.

cutpoints

(numeric())
Alternative to cuts if discretise is true, provide exact cutpoints for discretisation. cuts is ignored if cutpoints is non-NULL.

scheme

(character(1))
Method of discretisation, either "equidistant" (default) or "quantiles". See reticulate::py_help(pycox$models$LogisticHazard$label_transform).

cut_min

(integer(1))
Starting duration for discretisation, see reticulate::py_help(pycox$models$LogisticHazard$label_transform).

model

(character(1))
Corresponding pycox model.

Value

No return value.

Vectorised Logical requireNamespace

Description

Helper function for internal use. Vectorises the requireNamespace function and returns TRUE if all packages, x, are available and FALSE otherwise.

Usage

requireNamespaces(x)

Arguments

x

(character())
string naming the packages/name spaces to load.

Value

No return value.

Set seed in R numpy and torch

Description

To ensure consistent results, a seed has to be set in R using set.seed as usual but also in numpy and torch via reticulate. Therefore this function simplifies the process into one funciton.

Usage

set_seed(seed_R, seed_np = seed_R, seed_torch = seed_R)

Arguments

seed_R

(integer(1))
seed passed to set.seed.

seed_np

(integer(1))
seed passed to numpy$random$seed. Default is same as seed_R.

seed_torch

(integer(1))
seed passed to numpy$random$seed. Default is same as seed_R.

Value

No return value.

Simulate Survival Data

Description

Function for simulating survival data.

Usage

simsurvdata(n = 100, trt = 2, age = 2, sex = 1.5, cens = 0.3)

Arguments

n

(integer(1))
Number of samples

trt, age, sex

(numeric(1))
Coefficients for covariates.

cens

(numeric(1))
Proportion of censoring to be generated, cut-off time is then selected as the quantile that results in cens.

Details

Currently limited to three covariates, Weibull survival times, and Type I censoring. This will be expanded to a flexible simulation function in future updates. For now the function is primarily limited to helping function examples.

Value

data.frame()

Examples

simsurvdata()

Safely convert a survival matrix prediction to a relative risk

Description

Many methods can be used to reduce a discrete survival distribution prediction (i.e. matrix) to a relative risk / ranking prediction. Here we define the predicted relative risk as the sum of the predicted cumulative hazard function - which can be loosely interpreted as the expected number of deaths for patients with similar characteristics.

Usage

surv_to_risk(x)

Arguments

x

(matrix())
TxN survival matrix prediction where T is number of time-points and N is number of predicted observations. Colum names correspond to predicted time-points and should therefore be coercable to numeric and increasing. Entries are survival predictions and should be (non-strictly) decreasing in each row.

Value

A numeric vector with the expected number of deaths.

References

Sonabend, R., Bender, A., & Vollmer, S. (2021). Evaluation of survival distribution predictions with discrimination measures. http://arxiv.org/abs/2112.04828.

survivalmodels: Models for Survival Analysis

Description

Author(s)

See Also

Build a Keras Multilayer Perceptron

Description

Usage

Arguments

Details

Value

Build a Pytorch Multilayer Perceptron

Description

Usage

Arguments

Details

Value

Compute Concordance of survivalmodel Risk

Description

Usage

Arguments

Value

Examples

Cox-Time Survival Neural Network

Description

Usage

Arguments

Details

Value

References

DeepHit Survival Neural Network

Description

Usage

Arguments

Details

Value

References

DeepSurv Survival Neural Network

Description

Usage

Arguments

Details

Value

References

Get Keras Optimizer

Description

Usage

Arguments

Details

Value

Get Pytorch Activation Function

Description

Usage

Arguments

Details

Value

Get Torchtuples Callbacks

Description

Usage

Arguments

Value

Get Pytorch Weight Initialization Method

Description

Usage

Arguments

Details

Value

Get Pytorch Optimizer

Description

Usage

Arguments

Details

Value

Install Keras and Tensorflow

Description

Usage

Arguments

Value

Install Pycox With Reticulate

Description

Usage