Non-determinism

[ptheywood]

The model appears to be non-determinitic.

I.e. run the Ex3 solution branch multiple times and there will be minor differentces in the output CSV.

At a quick glance this should be deterministic as message iteration order should be stable between subsequent runs, and < ops are used for distance checks rather than <=.

It could just be weird floating point comparisons, but I would have expected that to be stable as is. Might we worth checking without fmadd / debug build.

It would probably be best to provide a deterministic model for the tutorial.