Mathematics · Book
Linear Algebra and Learning from Data
Strang begins with the four fundamental subspaces and the singular value decomposition, then turns to special matrices, computation with large matrices and compressed sensing.
Good fit if
Connects spectral theory, low-rank matrix approximations, and optimization directly to neural networks and data science (MIT 18.065).
About the book
Later chapters cover probability and statistics, optimization, the architecture of neural nets, stochastic gradient descent and backpropagation. The book presents a complete course in the linear algebra behind learning from data, written for MIT 18.065.
Author
Readers
Where next
Connections
6Its authors, the courses that use it and the lectures that cite it.

