R/learner_survivalmodels_surv_deephit.R

In mlr-org/mlr3extralearners: Extra Learners For mlr3

#' @title Survival DeepHit Learner
#' @author RaphaelS1
#' @name mlr_learners_surv.deephit
#'
#' @description
#' Neural network 'Deephit' for survival analysis.
#' Calls [survivalmodels::deephit()] from pacakge 'survivalmodels'.
#'
#' @template learner
#' @templateVar id surv.deephit
#'
#' @template install_survivalmodels
#'
#' @details
#' Custom nets can be used in this learner either using the
#'  [survivalmodels::build_pytorch_net] utility function or using `torch` via \CRANpkg{reticulate}.
#'  The number of output channels depends on the number of discretised time-points, i.e.
#'  the parameters `cuts` or `cutpoints`.
#'
#' @references
#' `r format_bib("lee2018deephit")`
#'
#' @template seealso_learner
#' @template example
#' @export
LearnerSurvDeephit = R6Class("LearnerSurvDeephit",
  inherit = mlr3proba::LearnerSurv,

  public = list(
    #' @description
    #' Creates a new instance of this [R6][R6::R6Class] class.
    initialize = function() {

      ps = ps(
        frac = p_dbl(default = 0, lower = 0, upper = 1, tags = "train"),
        cuts = p_int(default = 10L, lower = 1L, tags = "train"),
        cutpoints = p_uty(tags = "train"),
        scheme = p_fct(default = "equidistant", levels = c("equidistant", "quantiles"),
          tags = "train"),
        cut_min = p_dbl(default = 0, lower = 0, tags = "train"),
        num_nodes = p_uty(default = c(32L, 32L), tags = "train"),
        batch_norm = p_lgl(default = TRUE, tags = "train"),
        dropout = p_dbl(lower = 0, upper = 1, tags = "train"),
        activation = p_fct(default = "relu",
          levels = c("celu", "elu", "gelu", "glu", "hardshrink", "hardsigmoid", "hardswish",
            "hardtanh", "relu6", "leakyrelu", "logsigmoid", "logsoftmax", "prelu",
            "rrelu", "relu", "selu", "sigmoid", "softmax", "softmax2d", "softmin",
            "softplus", "softshrink", "softsign", "tanh", "tanhshrink", "threshold"),
          tags = "train"),
        custom_net = p_uty(tags = "train"),
        device = p_uty(tags = "train"),
        mod_alpha = p_dbl(default = 0.2, lower = 0, upper = 1, tags = "train"),
        sigma = p_dbl(default = 0.1, lower = 0, tags = "train"),
        optimizer = p_fct(default = "adam",
          levels = c("adadelta", "adagrad", "adam", "adamax", "adamw", "asgd", "rmsprop", "rprop",
            "sgd", "sparse_adam"), tags = "train"),
        rho = p_dbl(default = 0.9, tags = "train"),
        eps = p_dbl(default = 1e-8, tags = "train"),
        lr = p_dbl(default = 1, tags = "train"),
        weight_decay = p_dbl(default = 0, tags = "train"),
        learning_rate = p_dbl(default = 1e-2, tags = "train"),
        lr_decay = p_dbl(default = 0, tags = "train"),
        betas = p_uty(default = c(0.9, 0.999), tags = "train"),
        amsgrad = p_lgl(default = FALSE, tags = "train"),
        lambd = p_dbl(default = 1e-4, lower = 0, tags = "train"),
        alpha = p_dbl(default = 0.75, lower = 0, tags = "train"),
        t0 = p_dbl(default = 1e6, tags = "train"),
        momentum = p_dbl(default = 0, tags = "train"),
        centered = p_lgl(default = TRUE, tags = "train"),
        etas = p_uty(default = c(0.5, 1.2), tags = "train"),
        step_sizes = p_uty(default = c(1e-6, 50), tags = "train"),
        dampening = p_dbl(default = 0, tags = "train"),
        nesterov = p_lgl(default = FALSE, tags = "train"),
        batch_size = p_int(default = 256L, tags = c("train", "predict")),
        epochs = p_int(lower = 1L, upper = Inf, default = 1, tags = "train"),
        verbose = p_lgl(default = TRUE, tags = "train"),
        num_workers = p_int(default = 0L, tags = c("train", "predict", "threads")),
        shuffle = p_lgl(default = TRUE, tags = "train"),
        best_weights = p_lgl(default = FALSE, tags = "train"),
        early_stopping = p_lgl(default = FALSE, tags = "train"),
        min_delta = p_dbl(default = 0, tags = "train"),
        patience = p_int(default = 10, tags = "train"),
        interpolate = p_lgl(default = FALSE, tags = "predict"),
        inter_scheme = p_fct(default = "const_hazard",
          levels = c("const_hazard", "const_pdf"), tags = "predict"),
        sub = p_int(default = 10L, lower = 1L, tags = "predict")
      )

      ps$add_dep("rho", "optimizer", CondEqual$new("adadelta"))
      ps$add_dep("eps", "optimizer", CondAnyOf$new(c("adadelta", "adagrad", "adam", "adamax",
        "adamw", "rmsprop", "sparse_adam")))
      ps$add_dep("lr", "optimizer", CondEqual$new("adadelta"))
      ps$add_dep("weight_decay", "optimizer",
        CondAnyOf$new(c("adadelta", "adagrad", "adam", "adamax", "adamw",
          "asgd", "rmsprop", "sgd")))
      ps$add_dep("learning_rate", "optimizer",
        CondAnyOf$new(c("adagrad", "adam", "adamax", "adamw", "asgd", "rmsprop", "rprop",
          "sgd", "sparse_adam")))
      ps$add_dep("lr_decay", "optimizer", CondEqual$new("adadelta"))
      ps$add_dep("betas", "optimizer", CondAnyOf$new(c("adam", "adamax", "adamw", "sparse_adam")))
      ps$add_dep("amsgrad", "optimizer", CondAnyOf$new(c("adam", "adamw")))
      ps$add_dep("lambd", "optimizer", CondEqual$new("asgd"))
      ps$add_dep("t0", "optimizer", CondEqual$new("asgd"))
      ps$add_dep("momentum", "optimizer", CondAnyOf$new(c("sgd", "rmsprop")))
      ps$add_dep("centered", "optimizer", CondEqual$new("rmsprop"))
      ps$add_dep("etas", "optimizer", CondEqual$new("rprop"))
      ps$add_dep("step_sizes", "optimizer", CondEqual$new("rprop"))
      ps$add_dep("dampening", "optimizer", CondEqual$new("sgd"))
      ps$add_dep("nesterov", "optimizer", CondEqual$new("sgd"))

      ps$add_dep("min_delta", "early_stopping", CondEqual$new(TRUE))
      ps$add_dep("patience", "early_stopping", CondEqual$new(TRUE))

      ps$add_dep("sub", "interpolate", CondEqual$new(TRUE))
      ps$add_dep("inter_scheme", "interpolate", CondEqual$new(TRUE))

      super$initialize(
        id = "surv.deephit",
        feature_types = c("integer", "numeric"),
        predict_types = c("crank", "distr"),
        param_set = ps,
        man = "mlr3extralearners::mlr_learners_surv.deephit",
        packages = c("mlr3extralearners", "survivalmodels", "distr6", "reticulate"),
        label = "Neural Network"
      )
    }
  ),

  private = list(
    .train = function(task) {

      pars = self$param_set$get_values(tags = "train")
      invoke(
        survivalmodels::deephit,
        data = data.table::setDF(task$data()),
        time_variable = task$target_names[1L],
        status_variable = task$target_names[2L],
        .args = pars
      )

    },

    .predict = function(task) {

      pars = self$param_set$get_values(tags = "predict")
      newdata = ordered_features(task, self)

      pred = invoke(
        predict,
        self$model,
        newdata = newdata,
        distr6 = FALSE,
        type = "all",
        .args = pars
      )

      list(crank = pred$risk, distr = pred$surv)

    }
  )
)

.extralrns_dict$add("surv.deephit", LearnerSurvDeephit)