eureqa

Discuss how polywell fusion works; share theoretical questions and answers.

Moderators: tonybarry, MSimon

Post Reply
jcoady
Posts: 141
Joined: Fri Jul 15, 2011 4:36 pm

eureqa

Post by jcoady »

Eureqa is a software tool for detecting equations and hidden mathematical relationships in your data. Science news magazine published this article about it.

http://www.sciencenews.org/view/feature ... _Scientist

Someone posed this question on a eureqa forum:

"While reading about the remarkable abilities of the Eureqa software in Science News Magazine today, I wondered if getting Eureqa involved with fusion power generation experiments might help researchers achieve a sustained reaction. After working for decades toward that goal, maybe Eureqa can clarify where the problems lie. "

This answer was provided by Michael Schmidt one of the creators of Eureqa.

"Definitely, particularly in exploring what properties impact the various engineering challenges. You could test and analyze everything, or use Eureqa to identify what influences are evidenced from data much more quickly."

You can get more information on and download a free copy of Eureqa form here:

http://creativemachines.cornell.edu/eureqa

I decided to give it a try and feed it some data from my polywell PIC simulator. I output the potential well data from the polywell simulator and imported it into the eureqa program. After running for about a minute it produced this result for an equation of the potential well with the data that I fed it.

http://dl.dropbox.com/u/5095342/Polywell/eureqa.jpg

The experimental data that I fed it can be found here if you want to give it a try. From the Eureqa software application, just click File menu item and then click "Import Data...." and feed it a copy of this file.

http://dl.dropbox.com/u/5095342/Polywel ... l-well.txt

What do you think would be some good experiments to run with Eureqa?

jcoady
Posts: 141
Joined: Fri Jul 15, 2011 4:36 pm

Post by jcoady »

Here is a video from the creator of Eureqa explaining how they came up with it. I think the science news article above got some of its information from this video presentation.

http://science-education.pppl.gov/video ... H.264.html

You can also view the video using the QuckTime video player and using the following url in the video player.

http://www.pppl.gov/video/SOS29JAN2011_ ... pson-H.mov

happyjack27
Posts: 1439
Joined: Wed Jul 14, 2010 5:27 pm

Post by happyjack27 »

bah. they should try different basis for the solutions, rather than just polynominal. e.g. they should try at the very least trigonometric basis as well. and a laplace transformed one.

also it should fit data from runs w/different parameters. so you can find how the settings effect the pwell.

jcoady
Posts: 141
Joined: Fri Jul 15, 2011 4:36 pm

Post by jcoady »

You can configure Eureqa to use different basis for the solution such as trigonometric. Here is a short video tutorial of eureqa finding the equation for a swinging pendulum from some measurement data.

http://www.youtube.com/watch?v=NhC1Qb-PQ5Q

Part 2 of the tutorial finds the equation of a pendulum using second derivatives.

http://www.youtube.com/watch?v=k7GCqfV5 ... re=related

Aero
Posts: 1200
Joined: Mon Jun 30, 2008 4:36 am
Location: 92111

Post by Aero »

What do you think would be some good experiments to run with Eureqa?
Develop a formula for the masses of elementary particles. The Heim mass formula (Google it) gives very good values for the masses of many particles but it is a very complex formula and not a predictive formula. That is, it gives a lot of good results, and also results for which no particles are known.

I suggest that Heim's mass formula would be a place to start gathering the input data, then see if Eureqa could provide a less complex relationship between the data and the particle masses.
Aero

happyjack27
Posts: 1439
Joined: Wed Jul 14, 2010 5:27 pm

Post by happyjack27 »

What do you think would be some good experiments to run with Eureqa?
social-political stats?
e.g. get data on a bunch of social and political statistics on countries (e.g. funding levels, education levels, wealth distribution, etc...) and solve for a dependent variable. (e.g. GDP per capita). then you could say, hmm... if we raise spending on x and lower it on y, we can increase overall economic output). might help to get two years of data so you can compare time-lagging effects via finding year-to-year autocorrelations.

i would love to see google do an app like that - a factor analysis / linear optimization program. where you can try different things and publish your results w/your datasets and people can comment on results and vote on them as e.g. interesting / important, etc.

Post Reply