When dealing with large matrices, it helps to have an algorithm to break multiplication into operations on matrix subblocks. Tens of thousands of indices add up to gigabytes, and you’re going to have trouble calling X.matmul on a matrix library backend. Here’s a simple implementation that plays to the tune of Major->Minor->Major to reassemble the… Continue reading Block Matrix Multiplication
Category: Linear Algebra
RRQR Factorization and Spanning Boxes
The famous QR factorization algorithm can be BLAS-3 optimized. Using the Woodbury matrix identity, we implement a block-based Rank-Revealing QR and point out a connection between pivot selection and object detection.