During my student-research “office hours” today, Kate Storey-Fisher (NYU) told me about issues she is having doing certain kinds of numerical integrals for her project on two-point statistics. That got me fired up about adaptive sampling for integration. I wrote a code notebook that does an adaptive one-dimensional integration. As I hoped, the precision of the integral (in some limit, and in some cases) seems to improve exponentially with the number of samples, provided that they are placed very carefully.
No comments:
Post a Comment