Eureqa

from Wikipedia, the free encyclopedia
Eureqa
Basic data

developer Michael Schmidt ( Cornell University , NY) and Hod Lipson ( Columbia University , NY); Distributor: Nutonian, Inc.
Publishing year November 2009
operating system Windows, Linux, Mac OS
category Artificial intelligence
License proprietary
www.nutonian.com

Eureqa is the brand name of a computer program that was developed in the Artificial Intelligence Lab at Cornell University in 2007 and has been commercially available by Nutonian, Inc. since 2011 with additional functions and services. The Modeling - Engine of Eureqa uses evolutionary algorithms to deal with artificial intelligence to derive mathematical equations from comprehensive records in einfachster form, see Symbolic regression .

development

prehistory

Companies have been involved in data science since the 1970s . H. Extraction of information from large amounts of data by employing teams of scientists who performed predictive and statistical modeling with programs and programming languages ​​such as R , Python , SAS (Statistical Analysis System) and SQL, as well as with their own knowledge of mathematical and statistical relationships.

concept

In 2007, Michael Schmidt, at that time a doctoral student at Cornell University, Computational Biology department , assumed that it should be possible to use evolutionary algorithms to explain mathematical relationships in sets of experimental or statistical data. An investigation of this approach makes sense, especially under the aspects that both the complexity of the examined problems and the amount of data in research continue to increase.

Under his supervisor Hod Lipson , director of the Creative Machines Lab of Columbia University , developed Schmidt a working artificial intelligence modeling engine as a "virtual data scientist" ( virtual data scientists ) consisting of predetermined data automatically and time-saving predictive and analytical models created and makes this available to the (human) experts. The program does not work as a “ black box ”, but the individual modeling steps can be traced in situ and data can be added or removed at any time.

The name “Eureqa” came after Archimedes' alleged exclamation “ Eureka! "( Εὕρηκα ) ( I have [it] found ), the" k "has been replaced by a" q "to the association with the word equation ( Equation ) cause.

Evolutionary search method

The program applies random equations to the entered data. Most of the equations do not provide any further results, but some equations give better results in curve fitting than others. These equations then form the basis for a new process known as an “evolutionary search” (among billions of equations and equation elements), until finally an equation, the result, optimally correlates the data entered at the beginning.

In addition, the program learns and uses successful equation elements from previous searches to solve new problems. The program also rates the solutions found according to their simplicity, i. H. a less complex formula is preferred over a more complex formula if both formulas provide the same data fit.

marketing

  • At the beginning of November 2009 the program was released as freeware for download .
  • Since June 2011, Eureqa has been offered by Nutonian Inc. as a commercial package with additional programs and additional services.
  • "Eureqa" has been protected as a brand name since 2014 .

Application examples

According to Lipson, the utility of the software lies particularly in research areas in which a great deal of data is generated, but for which there is still no explanatory theory as to how these data are related. For this reason, Lipson asked scientists from various disciplines to make their results available in order to test the versatility of Eureqa.

physics

Simple harmonic oscillator

biology

  • In the October issue of Physical Biology (2011) Lipson then described an experiment with data on the rhythmic fluctuation of glucose ("glycolytic oscillation") and glucose-converting enzymes when yeast cells become anaerobic. Its Eureqa analysis provided seven well-known equations with defined constants. These constants were of interest to John Wikswo ( Vanderbilt University , Nashville ) and Jerry Jenkins (HudsonAlpha Institute for Biotechnology, Huntsville ), who supplied the data but had difficulty determining these constants. Wikswo referred to this research approach as "automated biology explorer" ("automatic biology explorer").
Bacillus subtilis , Gram stain. The oval, unstained structures are the endospores.
  • The biologist Gurol Suel ( University of Texas ) studies the growth of Bacillus subtilis . He had already developed a mathematical model with 16 variables that could describe the experimental data on concentrations of various proteins and subcellular factors related to the endospore formation of B. subtilis . In a collaboration with Lipson - and after an initial failure with the program - renewed modeling with Eureqa delivered a surprising result: A new equation with only 7 variables was able to relate the existing data to Suels and even provided consistent results with new data generated afterwards . But this interesting result is offset by the fact that so far neither Lipson nor Suel can conclusively explain the theory behind the new equation.

Other scientific subjects

reception

By 2012, 20,000 people and by 2015 more than 80,000 people, including researchers, students and Fortune 500 companies, downloaded and used the program and applied it to questions as diverse as the behavior of animals in cattle herds or the movement of stocks on the stock exchange .

In relation to Eureqa, the mathematician Steven Strogatz has suggested that mathematical results that are still theoretically inexplicable could be harbingers of an “end of insight” if computer programs are able to establish scientific relationships that people recognize as correct, but they do intellectually can no longer follow.

DaSilva argues in a similar direction. He emphasizes as a special feature “Eureqa's ability to 'discover' (ie rediscover) the laws of nature” and assumes that in the future human scientists, together with networks of evolutionary algorithmic programs, will carry out experiments and evaluations and find results that [in the Technology] can be used further without understanding the scientific principles behind it, on which these results are based.

Web links

Individual evidence

  1. ^ Gregory Piatetsky: Four main languages ​​for Analytics, Data Mining, Data Science . KDnuggets. Retrieved May 19, 2016.
  2. ↑ Michael Schmidt: Clarifying the uses of artificial intelligence in the enterprise . TechCrunch. Retrieved May 19, 2016.
  3. ↑ Dennis Keohane: Nutonian - At the Cutting Edge of Technology, Science, and Data Analysis . VentureFizz.
  4. ↑ a b Free translation
  5. ↑ Antonio Regalado: 35 Innovators Under 35 . In: MIT Technology Review . August 19, 2014. Retrieved May 19, 2016.
  6. ↑ Dan Woods: How Nutonian's Machine Intelligence Model Shows How People And Technology Can Be Partners , Forbes Tech, August 13/2015; accessed on August 8, 2016.
  7. ↑ a b c Asaf Shtull wedding ring: An Israeli professor's 'Eureqa' moment , Haaretz, February 3, 2012; accessed on August 4, 2016.
  8. ^ A b Kenneth Chang: Hal, Call Your Office: Computers That Act Like Physicists , New York Times , April 2, 2009; Retrieved August 3, 2016.
  9. ↑ Farhad Manjoo: Will Robots Steal Your Job? , Slate , September 30, 2009; accessed on August 4, 2016.
  10. ↑ a b Brandon Keim: Download Your Own Robot Scientist , Wired , December 3, 2009; accessed on August 4, 2016.
  11. ↑ CBIG Consulting Leverages Cutting-Edge Eureqa Software to Disrupt Analytic Marketplace and Expand Data Science Practice , prweb, February 19, 2014; accessed on August 9, 2016.
  12. ↑ Trade name applied for on March 13, 2013; Registration received on June 17th, 2014 ; accessed on August 8, 2016.
  13. ^ A b Rachel Ehrenberg: Software Scientist , Science News Digital, January 14, 2012; accessed on August 4, 2016.
  14. ↑ Michael Schmidt and Hod Lipson: Distilling Free-Form Natural Laws from Experimental Data , Science Vol. 323, No. 5924 (April 3, 2009), pp. 81-85; Retrieved August 3, 2016.
  15. ^ David Salisbury: Robot biologist solves complex problem from scratch , Vanderbilt University, October 13, 2011; accessed on August 6, 2016.
  16. ^ David Weinberger: Too Big to Know: Rethinking Knowledge Now That the Facts Aren't the Facts, Experts Are Everywhere, and the Smartest Person in the Room Is the Room . Basic Books, January 7, 2014, ISBN 978-0-465-08596-5 , p. 129.
  17. ↑ Literature list (selection) of publications in the field of astronomy which Eureqa was used to create.
  18. ↑ Literature list (selection) of publications in the field of chemistry , for which Eureqa was used.
  19. ↑ Literature list (selection) of publications in the field of mechanics / biomechanics , which Eureqa was used to create.
  20. ↑ Literature list (selection) of publications in the field of materials science , for which Eureqa was used.
  21. ↑ Literature list (selection) of publications in the field of computer science which Eureqa was used to create.
  22. ↑ Literature list (selection) of publications in the field of natural analogue optimization methods , for the creation of which Eureqa was used.
  23. ↑ Literature list (selection) of publications in the field of medicine which Eureqa was used to create.
  24. ↑ Literature list (selection) of publications in the field of neurology , which Eureqa was used to compile.
  25. ^ Nicholas Allgaier and Ryan McDevitt: Reverse Engineering the Brain with Eureqa (PDF); accessed on August 4, 2016.
  26. ↑ Literature list (selection) of publications in the field of psychology , for which Eureqa was used.
  27. ↑ Literature list (selection) of publications in the field of environmental sciences which Eureqa was used to create.
  28. ↑ Literature list (selection) of publications in the field of music that Eureqa was used to create.
  29. ↑ Literature list (selection) of publications in the field of general topics that Eureqa was used to compile.
  30. ↑ Statement by Nutonian.
  31. ↑ List of some companies that are customers of Nutonian Inc .; accessed on August 11, 2016.
  32. ↑ Free translation: "End of insight" or "End of the ability to understand".
  33. ^ Samuel Arbesmann: Explain It to Me Again, Computer. What if technology makes scientific discoveries that we can't understand? , Slate, February 25, 2013; accessed on August 10, 2016.
  34. ↑ Free translation: Eureqa's ability to 'discover' (i.e. rediscover) the laws of nature .
  35. ↑ Extropia DaSilva: Eureqa! Signs of the Singularity? , h Plus Magazine, March 25, 2011; accessed on August 6, 2016.