smi_sec_5_1_expectation: SMI on synthetic data, on expectation

In christianu7/aistats2020smi: Semi-Modular Inference

Description

Illustrate convenience of SMI using Monte Carlo integration over the data generation

Usage

smi_sec_5_1_expectation(
  out_dir,
  n_iter = 1000,
  n_new = 1000,
  phi = 0,
  theta = 1,
  n = 25,
  m = 50,
  sigma_z = 2,
  sigma_y = 1,
  sigma_phi = Inf,
  sigma_theta = 0.5,
  eta_all = seq(0, 1, 0.025)
)

Arguments

out_dir	Directory where outputs are stored
n_iter	number of iterations
n_new	number of iterations
phi	Generative (true) value assumed for phi
theta	Generative (true) value assumed for theta
n	Sample size for Z
m	Sample size for Y
sigma_z	Likelihood variance for Z
sigma_y	Likelihood variance for Y
sigma_phi	Prior std dev phi
sigma_theta	Prior std dev eta
eta_all	Degree of influence of the unbiased module into phi

christianu7/aistats2020smi documentation built on March 7, 2021, 2:40 p.m.