sim_learning: Simulates a learning model

In yukiyanai/rgamer: rgamer: Learn Game Theory Using R

Simulates a learning model

Description

sim_learning() simulates learning dynamics in a normal-form game expected by an experienced weighted attraction (EWA) model.

Usage

sim_learning(
  game,
  n_samples,
  n_periods,
  type = "EWA",
  lambda = 1,
  delta = 0.5,
  rho = 0.5,
  phi = 0.5,
  A1_init = 0,
  A2_init = 0,
  N_init = 0,
  plot_range_y = NULL
)

Arguments

game	An object of normal_form class defined by normal_form().
n_samples	A positive integer specifying the number of samples to be simulated.
n_periods	A positive integer specifying how many times the game is played within each sample.
type	A character string to tell which learning models should be simulated. The available options are "EWA", "reinforcement" (choice reinforcement), and "belief" (belief based model). "reinforcement" and "belief" are special cases of "EWA".
lambda	A positive real value representing the players' sensitivity to attraction values of strategies. As lambda gets larger, the choice will be dependent on attraction values more heavily. As lambda gets close to 0, a strategy will tend to be chosen randomly.
delta	A real number between 0 and 1. This parameter controls how fast attraction values of strategies that are not chosen are updated. If delta = 0, attraction is updated only for the strategy that is selected at the given period (i.e., reinforcement learning is implemented). If delta = 1, attraction is updated equally for all strategies (i.e., belief-based learning model is applied).
rho	A real value between 0 and 1. This parameter controls the learning speed. rho = 0 for "reinforcement" leaning and "belief" based learning.
phi	A real value between 0 and 1. This parameter controls how much attraction values at the current period are constrained by the past attraction values. If phi = 0, the past attraction values are ignored. phi = 1 for "reinforcement" leaning and "belief" based learning.
A1_init	An initial value of Player 1's attraction for each strategy.
A2_init	An initial value of Player 2's attraction for each strategy.
N_init	An initial value of N.
plot_range_y	Choose the range of vertical axis for plots. Available choices are "fixed", "full" and "free". If plot_range_y = "free", the range of y-axis depends on simulation results. If plot_range_y = "full", The range defined in game is used for each player, which can be different between players. With "fixed", the same y-axis is used for both players.

Details

Simulate plays of a normal-form game defined by normal_form() in a way expected by an EWA model.

Value

A list containing (1) a list of data frames of strategies chosen by each player, (2) a single long data frame of (1)'s data frames combined, (3) a list of each player's attraction values for each strategy (data frames), (4) a list of probability of each strategy being chosen (data frames), and (5) a plot of the simulation result (ggplot object).

Author(s)

Yoshio Kamijo and Yuki Yanai [e-mail]

yukiyanai/rgamer documentation built on June 14, 2024, 7:38 p.m.