Now I'm thinking about schemes where you iterate through lots of different permutations of visualizations, in at least two and three dimensions, and then someho

Now I'm thinking about schemes where you iterate through lots of different permutations of visualizations, in at least two and three dimensions, and then somehow calculate a score for how interesting the pattern is. Maybe using a sort of contrast algorithm like autofocus for a camera. If it looks noisy and random, then that's not an interesting visualizations.