Notes
Mostly about how the work actually gets done, and about the habits that make a result worth trusting.
Why I stopped storing deformations on a voxel lattice, and what became possible once I did.
Implicit representation papers all assert their field is continuous. I built the instruments to check, and the answer was more interesting than a yes.
If a continuous field only helps where sampling is the constraint, then the effect should scale with how badly sampled the data is. It does.
The most useful experiment I ran was the one designed to test whether my own contribution was doing anything.
Working on VISTA changed how I report results in a completely different field.
Signal adds coherently. Noise adds in quadrature. Everything else follows.
An essay I wrote away from the lab, about the questions that do not resolve. Hosted on Medium.