Regression for data science
This is a mini-course on regression, originally designed as part of the AI education program for TU Delft. I made it open for everyone who are insterested in learning regression in machine learning.
Clone the repository using git, or download it by click Code -> Download Zip at the top of this repository page.
Overview
Generally, We use regression to study relationships among dependent and independent variables presented in the data. When these regression models are used for predicting unseen input, the approach is categorized as supervised machine learning.
In this mini-course, you will learn the following topics:
- Simple linear regression (single input, single output, linear relationship, deterministic)
- Multiple linear regression (multiple inputs, single output, linear relationship, deterministic)
- Polynomial linear regression (multiple inputs, single output, non-linear relationship, deterministic)
- Logistic regression (continuous input(s), binary output, deterministic)
- Bayesian regression (regression using stochastic process)
In between these specific regression topics, there are a few regression-related topics:
- Regression metrics (how to evaluate your regression model?)
- Gradient descent algorithm (how to perform optimization numerically?)
- Bias variance trade-off (why you model under-fits or over-fits?)
- Regularization (how to mitigate the over-fitting problem?)
Materials for the lecture
The lectures are split into 11 short videos (edited to approximately 10 minutes on average). Feel free to use 1.25x or 1.5x speed to watch. Most topics are taught with the same dataset (about airplanes) that I have created for this lecture.
Each course video is accompanied by a Python Jupyter notebook in this repository.