Madhav-Kanda / pml_baselines

pml_baselines

Compare different baselines for the Baysian Inference problems.

We compare:

• MCMC- Markov Chain Monte Carlo
• ajax - Variational Inference using ajax.
• Laplace Approximation
• PyStan - Variational Inference using Stan

1. Coin toss Problem

Stan

coin_toss = """
data {
  int<lower=0> n;
  int<lower=0, upper=1> heads[n];
  real alpha0;
  real beta0;
}
parameters {
  real<lower=0, upper=1> theta;
}
model {
  heads ~ bernoulli(theta);
  theta ~ beta(alpha0, beta0);
}
"""

Model

Results

View the results here-https://anandshegde.github.io/pml_baselines/results/coin_toss/coin_toss_results.html

2. Linear Regression

View the results here-https://anandshegde.github.io/pml_baselines/results/linear_regression/linear_regression_results.html

3. Logistic Regression

View the results here-https://anandshegde.github.io/pml_baselines/results/logistic_regression/logistic_regression_results.html

4. Weibull Poisson

View the results here-https://anandshegde.github.io/pml_baselines/results/weibull_poisson/weibull_poisson_results.html