Linear Algebra and Learning from Data by Gilbert Strang

Gilbert Strang taught linear algebra to a generation through MIT’s OpenCourseWare, and this is the book he wrote once it became clear the subject’s centre of gravity had moved to machine learning. Linear Algebra and Learning from Data is not a revision of his classic textbook — it is a different book, aimed at what data actually demands.

What the book covers

• Highlights of linear algebra: the four fundamental subspaces, factorisations and rank
• Computations with large matrices, randomised linear algebra and low-rank approximation
• Low-rank and compressed sensing, matrix completion, and the SVD as the organising idea
• Special matrices, Fourier and circulant structure, and graph-based matrices
• Probability and statistics for data, covariance matrices and multivariate distributions
• Optimisation: gradient descent, stochastic gradient descent, and convexity
• Learning from data: the architecture of deep neural networks and backpropagation

Who it’s for

Graduate and advanced undergraduate students in data science, machine learning engineers who want deeper mathematical footing, and anyone who followed Strang’s lectures and wants the sequel.

Edition details

• Author: Gilbert Strang
• Publisher: Wellesley-Cambridge Press
• Format: Hardcover
• ISBN: 978-0692196380
• Language: English

Ships free within the United States from Shelfcove.