This week in class we discussed non-parametric tests like the U test.
I was working a lot on the project this week. I spent about a day reading up on MCMC techniques and time implementing t-SNE on the data to see if there are clusters that I might be interested in fitting.
I found a tricky bug in the integrator. The first step in the integrator was computing the time step, this requires the centered magnetic field and since I use that later in the algorithm I only wanted to compute it once and so saved it to a member variable. However since I was doing that before I computed the boundary conditions I was getting non-physical results on the first time step since the ghost cells hadn’t been set yet. Simply calling the boundary conditions function before the time step function resolved that issue. Now the single waves run stably, for a little bit at least. The issue looks like an error in the derivatives somewhere and I’m leaning towards the CT fields since I get similar results when forcing the Riemann solver into the L or R state.
The MHD contact wave