Chronology Current Month Current Thread Current Date
[Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

[Phys-L] Natural laws from experimental data2



[       From: Antti Savinainen via Phys-l <phys-l@mail.phys-l.org>
- To:phys-l@mail.phys-l.orgCc:Antti SavinainenSun, Aug 15 at 2:19 AMHi,

I suppose this is common knowledge in some circles, but I just found out
about this when preparing a lesson on technology, science & knowledge:
https://youtu.be/MSo6eeDsFlE

I find it astonishing that deep laws of physics (such as Hamiltonians)
could be found from data using symbolic regression algorithms, with no
predetermined physics laws. The paper was published in Science (2009).

It may be no surprise that the aforementioned premise has been challenged:
https://arxiv.org/abs/1210.7273

OTOH, the manuscript above was rejected by Science.

I wonder what is the best interpretation of the case.

Regards,

Antti Savinainen, PhD
Finland
_______________________________________________
Forum for Physics Educators
Phys-l@mail.phys-l.org
Phys-l Info Page
******************************************************************************************************************************
 In M. Schmidt, H. Lipson, Distilling free-form natural laws from experimental data, Science, 324 (2009),81–85,Schmidt described an engine to search for appropriate models for input data files. (Lipson was his supervisor.)
This app was soon commercialized by Nutonian as Eureqa which was available as a download (which can still be found.)   At length (2017) Nutonian was taken over by DataRobot.It is a random equation generator with evolutionary development using Genetic Algorithm methods.
In the version I used,  some preselection was available as to equation types: transcendental, exponential, etc., etc.It did not provide me with any Eureka moments sadly of the kind that Matlab offered.
 When neural networks later regained some momentum in dealing with very large data sets,  I was pleased to see that medical observables (for example) could be clustered into multi dimensional visual clouds so that a person without medical training could provide convincing diagnoses on this basis. Given the jumbo datasets needed, the clustering application is self organizing.