Data visualization

[Doug Coulter]

Just check this one out.


Now you could generate endless quibbles about this guys assumptions, normalizations (what does money mean in his context, is it inflation adjusted, adjusted to gold, adjusted to living expenses?) and so on, but this is a neat trick he's got that changes data into useful information if you know those things.

Troy Berg and I long ago wrote some code we called "fuzzy balls in space" that added another axis (3-d) and did more with the balls other than just size --we had shapes, textures, and probability distributions (the fuzzy part). Sadly, it was so specialized that it's not currently useful for fusors -- it was written to help a human being do things even neural nets have troubles with -- find hyperplane or simple shaped thresholds in N-space that separate classes of events from metrics we had taken on those events. Hoping that someday when he's not too busy (he was the real man on that one, I was just the idea guy) to generalize it so we can use it in our work.

I'm building up a similar picture in my head of how my fusor reacts to sweeps of all the parameters, and am taking data I hope to be able to someday present on that one. It's not a simple "more of this makes more of that" by any means, there are local peaks which a gradient following mechanism can't escape for example....and more than three parameters, so it's hard for most people to visualize. You could get most people to 5 dimensions + time with the code he did....after awhile you learn what's going to change as you rotate around some axis you can't see. Some people could do better -- a true case of computer-human synergy I'd like to see more push on myself.

Thanks BillF for finding this one and sending me a link!