Eureqa, the robot scientist (w/ Video)

December 7, 2009 by Lin Edwards, weblog

Eureqa screenshot. Image: Cornell Computational Synthesis Lab
( -- A new program, Eureqa, takes raw data and formulates scientific laws to suit, and it is available by free download to all scientists.

When the program first appeared in April this year, it was fed information on a double pendulum and in just a few hours it inferred Newton's second law of motion and the law of conservation of momentum from the data. Given other data, it could find laws that have so far eluded scientists.

Eureqa is a successor to robots that work out how to repair themselves, which were developed at the Computational Synthesis Lab at Cornell University by Dr Hod Lipson. The same algorithms that were used in the robots have been adapted for the analysis of any kind of data. These algorithms may help scientists find complicated equations and laws.

The program begins by examining the data for numbers that appear to be connected, and then suggests equations that fit the connections. Of the proposed equations most fail, but some are less wrong than others, and these are selected and modified and then repeatedly re-tested against the data and tweaked until a workable equation is identified.

In some cases there is not enough data to enable Eureqa to find equations, but in these cases the latest version of the program may identify the gaps in the data and even recommend experiments to supply the missing data.

Eureqa was able to calculate in hours equations that Newton took years to find, and Lipson hopes it can do the same for data such as the interactions between proteins, genomes and cell signals, which are so complicated that describing the interactions mathematically has so far been impossible. While Lipson envisaged the program as having application mainly in biological fields, it will analyze any data that can be presented in a .

Video: Cornell Computational Synthesis Lab

Dr John Wikswo of Vanderbilt University, who is using Eureqa to study the effects of cocaine on white blood cells, said that biology is far too complicated for humans to fully understand, but the Eureqa project may find solutions. Teamed with other gadgets developed by Lipson, Eureqa can adjust valves controlling the nutrients and toxins being fed to cells, and make changes faster than any human. Dr Wikswo said the program not only derives the equations, but also the experiments needed to come up with the equations.

Dr Wikswo explained that scientists usually work by keeping everything constant except one variable, but that works best for linear systems and not so well for biological systems, which are more complex, and which can only be understood fully by changing many variables. Understanding which variables to change and what the results mean can be incredibly complicated, but Eureqa should be able to help.

Eureqa was released in response to an overwhelming number of requests from scientists asking Lipson to analyze their data for them. The program is available for free download now, but is still being refined by Lipson and his colleague Michael Schmidt. One of the problems is its tendency to return suitable equations but with variables that are not understood. The equations work and make accurate predictions, and must be true, but no one can understand how they work. Lipson likens the situation to trying to explain the laws of energy conservation to mathematicians from medieval times, who did not have the vocabulary needed to understand the mathematics.

One example of this is the use of Eureqa by University of Texas Southwestern's Dr Gurol Suel to analyze data on cell division and growth. Eureqa developed equations, and although Dr Suel is not sure what they mean, he said the results are still useful, and can be used as a starting point for further work, and can help in the development of new hypotheses about the cells.

The next step is to devise algorithms to explain what Eureka is finding, possibly by relating the unknown concepts to those with which we are familiar. Meanwhile, the program is freely available for download at Cornell University's website.

More information: Eureqa page
via Wired
© 2009

Explore further: Being Isaac Newton: Computer derives natural laws from raw data

Related Stories

Mathematicians find new solutions to an ancient puzzle

March 14, 2008

Many people find complex math puzzling, including some mathematicians. Recently, mathematician Daniel J. Madden and retired physicist, Lee W. Jacobi, found solutions to a puzzle that has been around for centuries.

Researchers build a robot that can reproduce

May 11, 2005

One of the dreams of both science fiction writers and practical robot builders has been realized, at least on a simple level: Cornell University researchers have created a machine that can build copies of itself. Admittedly ...

New method for solving differential equations

January 24, 2008

Dutch-sponsored mathematician Valeriu Savcenco has developed new methods for the numerical solution of ordinary differential equations. These so-called multirate methods are highly efficient for large systems, where some ...

Quantum computing may actually be useful, after all

October 9, 2009

( -- In recent years, quantum computers have lost some of their luster. In the 1990s, it seemed that they might be able to solve a class of difficult but common problems — the so-called NP-complete problems ...

Recommended for you

Earth's deep mantle flows dynamically

March 25, 2019

As ancient ocean floors plunge over 1,000 km into the Earth's deep interior, they cause hot rock in the lower mantle to flow much more dynamically than previously thought, finds a new UCL-led study.

Scientists solve mystery shrouding oldest animal fossils

March 25, 2019

Scientists from The Australian National University (ANU) have discovered that 558 million-year-old Dickinsonia fossils do not reveal all of the features of the earliest known animals, which potentially had mouths and guts.


Adjust slider to filter visible comments by rank

Display comments: newest first

5 / 5 (1) Dec 07, 2009
While it delivers formulae it doesn't deliver the rationale for the formulae.
It may lead researchers to a symptomatic description of events but it won't get them to understand things (and let's face it: real world data on - as yet - not understood phenomena is a bit more tricky/noisy/prone to bias in measuring errors than pendulum swings are)

You can, mathematically, fit an infinite number of different formulae to a set of data points - but that doesn't tell you which fit makes sense.

The software could be a double edged sword: It could point out new correlations which scientists could study or it could create blinders by shunting enquiry down completely bollocksed avenues that look 'too good to be false'
...which this part
Dr Suel is not sure what they mean, he said the results are still useful,

seems to already demonstrate.
not rated yet Dec 16, 2009
I don't like the Microsoft minute in the beginning - after all, scientist rarely save their data in to excel worksheets. At least I don't -I save them into text files.
Secondly, it all looks nice, but really it doesn't tell you which fit makes sense physically.
And ultimately - it takes way too long to do those fits, don't you think. After all, this is a very simply law and it took it like 5 minutes to do the math. I can't even think how long would it take to fit something more complex.
5 / 5 (1) Dec 29, 2009
This s/w can perhaps be enhanced for use in any sound/time related fields of science, especially in medicine and health. It's quite revolutionary picking up patterns such as heartbeat and automating the patterns. So the medical scanner can entirely be based on bio rhythmic conditions. This must have been the scanner that star trek medics had been using to determine a health of a subject!

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.