rl_dnn_config: DNN Configuration for Reinforcement Learning

In RLescalation: Optimal Dose Escalation Using Deep Reinforcement Learning

DNN Configuration for Reinforcement Learning

Description

DNN (deep neural network) configuration for reinforcement learning. For detail, see Section 3.1 of the original paper.

Usage

rl_dnn_config(
  fcnet_hiddens = c(256L, 256L),
  fcnet_activation = c("relu", "tanh", "swish", "silu", "linear"),
  ...
)

Arguments

fcnet_hiddens	A positive integer vector. Numbers of units of the intermediate layers.
fcnet_activation	A character value specifying the activation function. Possible values are "ReLU" (default), "tanh", "Swish" (or "SiLU"), or "linear".
...	Other configurations. See source code of RLlib. https://github.com/ray-project/ray/blob/master/rllib/models/catalog.py

Value

A list of DNN configuration parameters

Examples

## Not run: 
escalation_rule <- learn_escalation_rule(
  J = 6, target = 0.25, epsilon = 0.04, delta = 0.1,
  N_total = 36, N_cohort = 3, seed = 123,
  rl_config = rl_config_set(
    iter = 1000, 
    # We change the DNN model
    model = rl_dnn_config(fcnet_hiddens = c(512L, 512L), fcnet_activation = "tanh")
  )
)
## End(Not run) 

RLescalation documentation built on April 4, 2025, 2:09 a.m.