loreloc/bayesian-gp-regression

Comparison of Several Bayesian Regression Techniques and Gaussian Processes

Bayesian and Gaussian Process regression (and some examples)

Comparison of several Regression techniques and Gaussian Process.

Contents

Examples taken from:

[1] MARTIN, OSVALDO. Bayesian Analysis with Python -: Implement Statistical Modeling and Probabilistic Programming Using pymc3. PACKT Publishing Limited, 2018.

Examples of classic Bayesian models:

• Linear regression
  • Robust linear regression
• Logistic regression
• Multivariate Linear and Logistic regression
• Poisson regression (ZIP)
• Polynomial regression (univariate and multivariate)
• Linear splines

Examples of Gaussian Processes:

• Gaussian Process Regression
• Regression with spatial autocorrelation
• Gaussian Process Classification
  • GP Classification with a More Complex Target
• Poisson Process (and Cox Process)

Application on real data: Yield crop prediction

• Preprocessing
• Bayesian models
• Gaussian processes (potato)
• Gaussian processes (wheat)

Repo structure

.
├── data                    # Datasets used in notebooks
├── world                   # Geodata for plotting world maps
├── 1.* ... 5.*.ipynb       # Notebooks
├── guassian_processes.py   # GP utility class
├── utils.py                # Utility functions
├── LICENSE
└── README.md

Installation

Recommended to create a venv; recommended to install pymc separatly with conda if on windows (Instructions);

then:

pip install -r requirements.txt

Usage

Just use them as a regular ipy notebooks