+- MKLab (https://mklab.gr)
+-- Forum: [INDEX] (https://mklab.gr/forumdisplay.php?fid=1)
+--- Forum: MATHEMATICS (https://mklab.gr/forumdisplay.php?fid=3)
+---- Forum: BOOKS (https://mklab.gr/forumdisplay.php?fid=6)
+----- Forum: NEW BOOKS (https://mklab.gr/forumdisplay.php?fid=42)
+------ Forum: FOREIGN (https://mklab.gr/forumdisplay.php?fid=91)
+------- Forum: PURE AND APPLIED MATHS (https://mklab.gr/forumdisplay.php?fid=94)
+-------- Forum: LINEAR ALGEBRA (https://mklab.gr/forumdisplay.php?fid=193)
+-------- Thread: Linear Algebra and Learning from Data [Strang] (/showthread.php?tid=2124)
Linear Algebra and Learning from Data [Strang] - mklabgr - 09-23-2026
Gilbert Strang’s Linear Algebra and Learning from Data presents linear algebra as a central mathematical foundation for modern data science and deep learning. The book begins with core concepts such as singular values, least squares, matrix factorizations, and low-rank approximation, showing how these techniques can extract the most important information from large datasets. It emphasizes practical matrix methods, including randomized algorithms for handling very large matrices, and connects the theory to problems encountered in data reduction and interpretation.
The book then moves from linear algebra to machine learning, explaining how the training of neural networks becomes a large-scale optimization problem solved through gradient descent and stochastic gradient descent. Strang develops the mathematics behind fully connected neural networks and convolutional neural networks, illustrating how these architectures use matrices and optimization to discover patterns in data.
Overall, the book is designed as an accessible bridge between a second course in linear algebra and the mathematical ideas underlying contemporary deep learning.
BOOK