SoylentNews is people
SoylentNews
Submission Preview
Graph Isomorphism Strikes Back
Story automatically generated by StoryBot Version 0.2.2 rel Testing.
Storybot ('Arthur T Knackerbracket') has been converted to Python3
Note: This is the complete story and will need further editing. It may also be covered
by Copyright and thus should be acknowledged and quoted rather than printed in its entirety.
FeedSource: [HackerNews]
Time: 2017-01-06 04:38:35 UTC
Original URL: https://www.quantamagazine.org/20170105-graph-isomorphism-retraction/ [quantamagazine.org] using UTF-8 encoding.
Title: Graph Isomorphism Strikes Back
--- --- --- --- --- --- --- Entire Story Below --- --- --- --- --- --- ---
Graph Isomorphism Strikes Back
Arthur T Knackerbracket has found the following story [quantamagazine.org]:
The legendary graph isomorphism problem may be harder than a 2015 result seemed to suggest.
January 5, 2017
The theoretical computer scientist László Babai [uchicago.edu] has retracted [uchicago.edu] a claim that amazed the computer science community when he made it just over a year ago. In November 2015, he announced that he had come up with a “quasi-polynomial” algorithm for graph isomorphism [quantamagazine.org], one of the most famous problems in theoretical computer science. While Babai’s result has not collapsed completely — computer scientists still consider it a breakthrough — its central claim has been found, after a year of close scrutiny, to contain a subtle error.
“In Laci Babai, you have one of the most legendary and fearsome theoretical computer scientists there ever was, and in graph isomorphism, one of the most legendary and fearsome problems,” wrote Scott Aaronson [scottaaronson.com], a theoretical computer scientist at the University of Texas, Austin, in an email. “A year ago, Laci threw an unbelievable knockout punch at [graph isomorphism], and now the problem itself seems to have gotten off the mat and thrown a counterpunch.”
The graph isomorphism problem asks for an algorithm that can spot whether two graphs — networks of nodes and edges — are the same graph in disguise. For decades, this problem has occupied a special status in computer science as one of just a few naturally occurring problems whose difficulty level is hard to pin down.
Roughly speaking, most computer science problems fall into one of two broad categories. There are “easy” problems, the ones that can be solved in a polynomial number of steps — if the size of the problem is denoted by n, the number of steps grows as, for example, n 2 or n 3. These problems can (generally) be solved efficiently on a computer. And there are “hard” problems, for which the best known algorithm takes an exponential (such as 2 n ) number of steps — far too many for a computer to carry out efficiently. Only a handful of natural problems, including graph isomorphism, seem to defy this dichotomy; computer scientists have struggled for decades to figure out just where graph isomorphism belongs.
Babai, a professor at the University of Chicago, had presented in late 2015 what he said was a “quasi-polynomial” algorithm for graph isomorphism. His work appeared to place the problem, if not firmly in the easy zone, then at least in its suburbs. But on January 4 he announced that while his algorithm still works (with some small tweaks) and has now been carefully checked by other computer scientists, it doesn’t run as fast as he had thought. It is “sub-exponential,” which moves the problem back into the suburbs of the hard zone.
Babai’s algorithm is, nevertheless, significantly faster than the previous best algorithm for graph isomorphism, which had held its title unchallenged for more than 30 years. “It’s still a massive improvement over the previous state of the art,” said Aaronson by email. Computer scientists conversant in Babai’s approach will presumably try to figure out whether further improvements can be milked from it, Aaronson predicted.
In a statement on his University of Chicago web page, Babai thanked Harald Helfgott [imj-prg.fr], a mathematician at the University of Göttingen and France’s National Center for Scientific Research, for “spotting this error and for spending months studying the paper in full detail.” Helfgott, in a blog post [wordpress.com], wrote that despite the error in Babai’s work, the remainder of his paper is “rich in innovative ideas.”
Erica Klarreich has been writing about mathematics and science for more than a decade. She has a doctorate in mathematics from Stony Brook University and is a graduate of the Science Communication Program at the University of California, Santa Cruz. Her work has been reprinted in “The Best Writing on Mathematics 2010” and “The Best Writing on Mathematics 2011.”
Leave a Comment
Your email address will not be published. Your name will appear near your comment.
Quanta Magazine moderates all comments with the goal of facilitating an informed, substantive, civil conversation about the research developments we cover. Comments that are abusive, profane, self-promotional, misleading, incoherent or off-topic will be rejected. We can only accept comments that are written in English.
Can you turn a two-dimensional fractal into a 3-D object? Break out your scissors and tape for a chance to win a 3-D printed sculpture.
A mission to collect samples from the far side of the moon could answer questions about a barrage of asteroids nearly 4 billion years ago.
The effort to unify quantum mechanics and general relativity means reconciling totally different notions of time.
How do physicists reconstruct what really happened in a particle collision? Through calculations that are so challenging that, in some cases, they simply can’t be done. Yet.
The same problem that caused the 2007 financial crisis also tripped up the polling data ahead of this year’s presidential election.
Abstractions navigates promising ideas in
science and mathematics. Journey with us
and join the conversation.
Back to top
-- submitted from IRC
