Lookup function to get value of upper and lower bounds for the current state

The current state of the multi armed bandit is marked based on the indices in the list of each combination of column and lower and upper bound type.

Usage

envir_to_bounds(current_envir, envir, interest_cols)

update_bounds(current_envir, envir, actions, selected_action)

create_anchor_inst(bounds, interest_cols)

generate_cutpoints(dataset, instance_id, interest_columns)

make_perturb_distn(n, interest_cols, dataset, instance_id, seed = 123)

Arguments

current_envir: List of indexes, the current state
envir: The current environment
interest_cols: Columns from the dataset that are of interest
selected_action: The index of selected action from the actions list
bounds: The upper and lower bounds of each column of interest
dataset: The dataset used to make the anchors
instance_id: The index of the target observation in the dataset
n: Number of samples
seed: Numerical. Seed to ensure that the perturbation distribution remains consistent.
interest_coluns: The columns of dataset to consider when creating lower and upper bounds

Value

A tibble of 1 x (2*p) where p is the number of columns of interest

New bounds

an instance of the anchor class

A list. Contains lower and upper bounds for each specific column of interest.

A purrr partial function that takes in N and returns N number of sample points around the instance