MIT develops algorithms to "solve" non-linear systems as sum of squares

[ashley]

http://web.mit.edu/newsoffice/2010/parrilo-convergence.html

[ssp]

His publications are here: http://www.mit.edu/~parrilo/pubs/index.html. This one in particular: http://arxiv.org/pdf/0909.0808v1 looks like it might be related.

[chancho → ssp]

I know this post is about Parrilo but I would just like to point out that Jesus De Loera (co-author of linked paper) is awesome.

Favorite paper: minimal tetrahedralization of a convex polyhedron is NP-hard. http://arxiv.org/PS_cache/math/pdf/0012/0012177v1.pdf

http://www.math.ucdavis.edu/~deloera/RECENT_WORK/recent.html

[gtt → ssp]

Can't open file from http://arxiv.org/pdf/0909.0808v1 evince says "Permission Denied". Could someone provide another link? Thanks in advance.

[esheldon → gtt]

Its just a PDF file, but the name lacks an extension. You might try downloading it, renaming it with a .pdf extension, and trying again to open it.

[nkh]

Does anyone have a link to the actual algorithms?

[KonaB → nkh]

I suppose the article is referring to the SOSTOOLS (where SOS = Sum of Squares) free MATLAB toolbox:

http://www.cds.caltech.edu/sostools

http://www.mit.edu/~parrilo/sostools

The article could have mentioned that more people have worked on this. The whole thing started while Parrilo and Prajna were Prof. John Doyle's grad students at Caltech / CDS. Yet once again, MIT's shameless self-promotion fails to give credit where credit is due...

[iamwil → KonaB]

I always hate how these articles never have links to the actual papers or references. I suppose most people don't care, but then again, this is MIT news. Nerds will read it. HN's the only place where you get links to stuff. Thanks.

[dschobel → iamwil]

http://web.mit.edu/newsoffice/ is the online alumni magazine whose, like every other alumni magazine on earth's, sole purpose is PR about the great things the school is doing.

It isn't exactly scientific american.

[iamwil → dschobel]

while true, I don't think it hurts for journalists of any sort to quote or link to their sources. And considering the audience of MIT alumni, I'm sure they'd appreciate it as much as I would.