femton

Kinetic triangulation problems are formulated similar to this. I've written an algorithm for a specific kind of triangulation schema [1] where collision events are sorted as quadratic inequalities. The math becomes quite involved when robustness is required and floating point representation is avoided. You have to treat the irrational time points of collision events as existing but expressed implicitly, as in a sort of unevaluated symbolic form. Kind of like imaginary numbers. [1] https://github.com/mpihlstrom/femton

The opaque nature of all the machine learning today has made me appreciate analytic algorithms more. For a number of years I've been exploring 2D representation using triangulations [1]. Working with analytic algorithms is like playing around with synthesizers whereas ML is more like using a sampler. [1] https://github.com/mpihlstrom/femton

This project is a first principles approach to image representation and manipulation. I got interested in alternatives to the pixel grid, intrigued by the in some respect even simpler division of 2D space by the simplicial complex, i.e., the triangulation. The main concrete contribution here is a triangulation algorithm which is specifically retains contour.