Tutorials Data Science Basics

Melissa K (Oct 2016)

The following tutorials are a step-by-step practical implementation of mathematical concepts that are at the very basis of machine learning. I believe that mastering linear least squares, simple gradient descent and singular value decomposition somewhat by hand is crucial to understanding more advanced machine learning concepts, including Deep Learning.

• The primary focus here is on practicality (Do it in Python) rather than deriving mathematical equations (there are tons of great tutorials out there that cover that part in detail).

• The secondary focus is on connecting the dots between basic mathematical concepts - for example you can fit a plane to a data cloud using either least squares method or singular value decomposition...you will learn the difference and implications.

• Finally, the solutions derived by hand will be compared to sklearn.

Table of content

1. Linear Regression Models - Least Squares Method

   Covering various multiple linear regression cases (e.g. continuous vs. mixed type features/independent variables to predict a continuous target/dependent variable using a closed-form solution)

2. Binomial Logistic Regression - Gradient Descent Optimization

   Covering simple gradient descent optimization as iterative method to predict a binary target/dependent variable.

3. Principal Component Analysis (PCA)

   Introducing Principal Component Analysis (PCA) using singular value decomposition to fit a plane to a data cloud in 3D.