They're nonuniformly-distributed, but that's' not necessarily bad.
They're currently all normalized, but we can probably skip that step and still get good results. The problem there is that the range of noise values is no longer easy to derive, so there would be some aesthetic fudging (which probably doesn't scale correctly with the number of dimensions).
They're nonuniformly-distributed, but that's' not necessarily bad.
They're currently all normalized, but we can probably skip that step and still get good results. The problem there is that the range of noise values is no longer easy to derive, so there would be some aesthetic fudging (which probably doesn't scale correctly with the number of dimensions).