Overview
Numerical computing is where clean mathematics meets a finite machine. An eigenvector becomes a loop that has to converge, a linear system becomes a matrix too large to invert, and a photograph becomes a problem in linear algebra.
Five projects at USI Lugano, all in MATLAB, all implemented rather than imported: the power method behind PageRank, spectral and inertial partitioning of meshes, a hand-written conjugate gradient that deblurs an image, the simplex method, and the same linear algebra turned on a network of people. One of them I ran on my own crawl of the Italian athletics web, just to see who came out on top.
What I built
- The PageRank algorithm, built on the power method.
- Spectral graph partitioning and an analysis of social networks.
- Conjugate gradient applied to image deblurring.
- Linear programming with the simplex method.
The five projects
Read-only · outputs from the MATLAB projects and their reports
Methods implemented rather than imported, then pointed at a problem where they have to earn their keep — the web, a mesh, a blurred photograph, a linear program, a social network.