Which is, really, all that this thing is: http://www.wired.com/wiredscience/2009/12/download-robot-scientist/

Still, it's a pretty cool thing, and maybe now that anyone can download and use it we'll see some interesting discoveries by people messing around with it. I wonder how powerful the software is, and whether it's limited to discovering scientific laws or perhaps extendable to, say, coming up with trends in the stock market.

That is cool.

That does seem pretty cool. Obviously, you can do the same thing with Matlab, Mathmatica, or Origin but they have pretty expensive licences and this thing is free. It's a pretty simple layout and has some clever fitting algorithms. If that's all you need, then this program is amazing.

That's Stargate SG1 stuff right there. And by that I mean a bad line from a scifi show, but it's true.

I hate to be a party pooper, but this program isn't actually anything revolutionary. Do you think scientists have just been running around with raw data and not been able to analyse it until now? We've had Mathematica since the 80s, which does exactly this but requires some basic coding.

This program is cool because it's very easy and simple, but we've had curve fitting algorithms for a long time.

"An ongoing challenge is the tendency of Eureqa to return equations that fit data, but refer to variables that are not yet understood. Lipson likened this to what would happen if time-traveling scientists presented the laws of energy conservation to medieval mathematicians."

Anyone else think of the Hitchhiker's Guide when reading that?

The answer to life is 42! But how the hell did it come up with that?

Ok, so I'm using Eureqa to analyse some data from a UV spectrum analysis of CdS quantum dots. This curve can tell you the diameter of these CdS nanoparticles. So, I've set this program running to tell me the equation describing the curve. It's been going for 5 hours, and.. well,


The green line is the curve I'm analysing. The pink line tells me how far Eureqa has analysed. As you can see, after 5 hours, it's barely gotten beyond the first peak. And the equation it's come up with is fairly retarded



0.00 x x^(15.69)

Er? That's equal to, er, 0. In conclusion, I'm not that impressed with Eureqa.

In case climate change skeptics get on to me and demand the raw data, it's here.

Mort-Hog, why are you trying to "hide the decline" in the diameter of CdS nanoparticles?

I'm not! It's just a mathematical trick I'm using to compare CdS nanoparticle diameter with that obtained from tree ring samples!

Did you leave it running? Has it found a more worthwhile equation yet?

Ok, it's now been running 13 hours. It's made much more progress



The pink line is just beyond the second peak. This is good, I need only a little bit more before I'm happy with this characterisation (nothing much interesting occurs beyond about the 250nm point). And the equation describing it is quite nice too



I'd probably just take the power to 0.99 as just 1, but other than that this equation looks pretty manageable. I just need to differentiate this now.

However, I have 25 of these graphs to analyse and this method would take over a week.

I can see its uses in certain circumstances. I might use this for some crystallography calculations, as you can use the dispersion relation of electrons travelling through a crystal to work out the vectors describing the lattice.

I would like to understand that. Knowing I never will is depressing.

So basically, it's a program that randomly generates possible equations from a list of data and variables.

About as effective as a brute force cracker..

Wow, if it takes that long I can see why they've provided the software for running it on a cluster of computers.

shut up shut up shut up SHUT UP

I'm staying up late working on a solid-state physics term paper, I came here to escape that stuff for a brief while

On a more serious note, I've seen a few articles on that piece of software and was pretty impressed. I'm kinda disappointed that it's just another curve-fitter. I should have learned my lesson about how sensationalized media coverage of science is, but I guess I haven't.

Mort, what's the equation you're looking for?
(And what are you working on? I like quantum dots)