みねん@人生(m.philos.y)

"Mathematical Pathways to Machine Learning" is a new introduction to the mathematical foundations of machine learning, published in August 2026. It is a collection of lecture notes that develops the mathematics from first principles and covers vectors and matrices, inner and outer products, orthogonality and projections, Gram–Schmidt and QR factorization, eigenvalues and eigenvectors, SVD, PCA, derivatives and gradients, Jacobians, Hessians, Taylor approximations, convexity, least squares, gradient descent, convergence and conditioning, Newton's method, momentum, Lagrange multipliers, KKT conditions, duality, and support vector machines. The mathematical concepts are directly connected to actual machine learning applications, including neural networks and backpropagation, PageRank, recommender systems, dimensionality reduction, regression, optimization, and SVMs. The presentation is accessible and very well structured, with worked examples and exercises throughout, including exam-practice problems and small computational exercises. zenodo.org/records/21752763