Video by freeCodeCamp.org · taught by Ayush Singh · streamed from YouTube. All rights remain with the creator.
Lesson 2: Machine Learning Algorithms in Depth
About this lesson
This long course by Ayush Singh teaches the classic machine learning algorithms in depth, explaining the mathematics behind each one as well as how to code it. It begins with the basics: what machine learning is, where it is used, its main types, the workflow of a typical project, the common challenges, how data is divided for training and testing, and how to recognise and fix underfitting and overfitting. Linear regression follows, from a visual explanation to the hypothesis function, the normal equation and its code, the cost function and gradient descent, ending with its assumptions and the regularised ridge and lasso variants. Logistic regression is then covered in the same way. A long section on support vector machines explains hard and soft margins, polynomial and radial basis function kernels, the primal and dual problems, sub-gradient and coordinate descent, transductive SVMs and support vector regression. Principal component analysis is built up from eigenvectors through standardisation, the covariance matrix and the cumulative energy of each component. A short section on learning theory explains the bias and variance trade-off and empirical risk minimisation. Decision trees and random forests come next, with entropy, information gain and Gini impurity, followed by bagging, boosting, AdaBoost, XGBoost and stacking. After Naive Bayes, nearest neighbours and clustering, the course ends with three complete projects. Afterwards, read Chapter 2 of the course book.