Stories
Slash Boxes
Comments

SoylentNews is people

SoylentNews is powered by your submissions, so send in your scoop. Only 19 submissions in the queue.
posted by janrinok on Friday March 14 2014, @03:12AM   Printer-friendly
from the here-it-is-over-there-somewhere dept.

Taco Cowboy writes:

"An AlJazeera article claims that Bayes' Theorem could be used to help with the search for the missing flight MH370.

Days after a Malaysian airliner with 239 people aboard went missing en route to Beijing, searchers are still struggling to find any confirmed sign of the plane. Authorities have acknowledged that they didn't even know what direction it was heading when it disappeared. In 2009, Air France Flight 447 en route to Paris from Rio de Janeiro vanished over the Atlantic Ocean, triggering the most expensive and exhaustive search effort ever conducted for a plane. After two years, officials could only narrow the location of the plane's black box down to an area the size of Switzerland. What took two years for other experts in the search for the black box, took only five days for consultants who applied the Bayes' Theorem, to finally find the device 12,000 feet under water. Read more inside.

'It's a very short, simple equation that says you can start out with hypothesis about something - and it doesn't matter how good the hypothesis is,' said Sharon Bertsch McGrayne, author of The Theory That Would Not Die: How Bayes' Rule Cracked the Enigma Code, Hunted Down Russian Submarines, and Emerged Triumphant from Two Centuries of Controversy.

The hypothesis is subject to change, based on probability, but can still be used with the theorem. Pretty much based on the concept of learning from experience, one can say. It is because of this character of the formula - forcing researchers to change their hypothesis with each new information - that the probability becomes more accurate.

Bayes' Theorem, which is also used in Google's driverless cars and predictions in stock markets, is based on probability. Because the theorem starts with a hypothesis - something McGrayne said 'can be very subjective' - it had been seen as controversial until the 1960s. But because it forces researchers to change their hypothesis with each new piece of information, the probability becomes more accurate.

The theorem was used in World War II to locate German U-boats and the lost nuclear submarine U.S.S. Scorpion. It was also used during the Cold War to spot Soviet submarines.

'The AF 447 search is rooted in Bayesian inference,' Lawrence D. Stone, chief scientist at Virginia-based scientific consultancy Metron - which was contacted to apply Bayes' Theorem in the search for the Air France plane - wrote in ORMS Today magazine in 2011. Bayes' Theorem 'allows the organization of available data with associated uncertainties and computation of the PDF (probability distribution function) for target location given these data,' he said.

Despite assistance from Australia, China, Thailand, Indonesia, Singapore, Vietnam, Philippines and the United States, Malaysian search efforts are even further from locating Flight MH370. The search area has been expanded to almost 27,000 square nautical miles - an area roughly equivalent to the state of Indiana - authorities said. That's more than 10,000 nautical square miles larger than the search for Air France Flight 447, before Bayes' Theorem was applied. Stone told Al Jazeera that in the current search for flight MH370, it is "highly unlikely" that Bayes' Theorem is being applied. That is not to suggest it is totally absent. Bayes' Theorem is pervasive, and those involved in the current search have applied a certain Bayesian flavour in their search, "but it then got upset when their prior calculations were incorrect," said statistician Professor Bradley Efron of Stanford University, as quoted by Al Jazeera, referring to the conclusion by Malaysian authorities that the MAS plane could have ended up in the Strait of Malacca.

Bayes' Theorem, after all, is all about learning from experience, which is probably why Efron said one would need "reasonably accurate past experiences" for the theorem to work. In other words, to calculate accurately to locate the plane."

This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 3, Interesting) by siwelwerd on Friday March 14 2014, @03:39AM

    by siwelwerd (946) on Friday March 14 2014, @03:39AM (#16169)

    Really 'editors', you couldn't catch that mispelling?

    • (Score: 4, Funny) by buswolley on Friday March 14 2014, @04:19AM

      by buswolley (848) on Friday March 14 2014, @04:19AM (#16175)

      Its a new element that powers science

      --
      subicular junctures
    • (Score: 4, Funny) by beckett on Friday March 14 2014, @04:41AM

      by beckett (1115) on Friday March 14 2014, @04:41AM (#16178)

      'theorum' is a perfectly cromulent word.

      • (Score: 1) by iWantToKeepAnon on Friday March 14 2014, @03:47PM

        by iWantToKeepAnon (686) on Friday March 14 2014, @03:47PM (#16461) Homepage Journal
        The spirit of 'theorum's embiggens the smallest man!
        --
        "Happy families are all alike; every unhappy family is unhappy in its own way." -- Anna Karenina by Leo Tolstoy
  • (Score: 5, Informative) by wjwlsn on Friday March 14 2014, @04:11AM

    by wjwlsn (171) on Friday March 14 2014, @04:11AM (#16174) Homepage Journal

    This is one of the better intros to Bayes' Theorem that I've seen, and it's mercifully short.
    http://advat.blogspot.ca/2012/04/intuitive-explana tion-of-bayes-theorem.html [blogspot.ca]

    If you make it through that and want more, I recommend this much longer (but still introductory) tutorial.
    http://yudkowsky.net/rational/bayes [yudkowsky.net]

    Application of Baye's Theorem to the search for the missing plane is left as an exercise for the student. :)

    --
    I am a traveler of both time and space. Duh.
    • (Score: 2) by c0lo on Friday March 14 2014, @03:59PM

      by c0lo (156) Subscriber Badge on Friday March 14 2014, @03:59PM (#16470) Journal

      Application of Baye's Theorem to the search for the missing plane is left as an exercise for the student. :)

      After repeatedly applying Bayes' theorem, I reached the conclusion that the probability to find what you are looking for in the last place you search is exactly 1.
      (grin)

      --
      https://www.youtube.com/watch?v=aoFiw2jMy-0 https://soylentnews.org/~MichaelDavidCrawford
      • (Score: 1) by egcagrac0 on Friday March 14 2014, @04:45PM

        by egcagrac0 (2705) on Friday March 14 2014, @04:45PM (#16499)

        Not if you stop looking before you find it.

  • (Score: 5, Insightful) by Daniel Dvorkin on Friday March 14 2014, @05:02AM

    by Daniel Dvorkin (1099) on Friday March 14 2014, @05:02AM (#16182) Journal

    ... we're in the 21st century, past time to stop acting like Bayes' Theorem is some startling revelation, and long past time for Bayesians to stop acting like they're Bold Rebels Speaking Truth To Power.

    It's a mathematical tool. If it's useful, great, use it. But please stop talking about it like it's Secret Wisdom known only to a Select Few, or a magic wand that can be waved at any problem to make it instantly solvable.

    --
    Pipedot [pipedot.org]:Soylent [soylentnews.org]::BSD:Linux
    • (Score: 1) by threedigits on Friday March 14 2014, @07:40AM

      by threedigits (607) on Friday March 14 2014, @07:40AM (#16203)

      Is Bayes' Theorem the new Fibonacci sequence?

    • (Score: 5, Insightful) by clone141166 on Friday March 14 2014, @09:37AM

      by clone141166 (59) on Friday March 14 2014, @09:37AM (#16248)

      With today's education standards, knowing even slightly advanced mathematics IS like being a magician.

    • (Score: 0) by Anonymous Coward on Friday March 14 2014, @03:21PM

      by Anonymous Coward on Friday March 14 2014, @03:21PM (#16432)

      It's basically how we normally think in every day life. The wikipedia article explains it reasonably. If someone told me they talked to someone on the subway (with no other facts to go by other than the assumption that half the population is composed of males and half of females) I can assume with, say, a 50% probability they were talking to a male and 50% that they were talking to a female. Now lets say that I learn that people who travel on subways are more likely to be male (the odds are 60% to 40%) because females are more afraid of traveling on subways, especially alone. Then that will change my perception of the probability of the gender of the person you were likely talking to.

      This is done in marketing all the time. New facts come in, you change your probability functions accordingly. This is basic intuition, as new facts come in you draw new conclusions and use those facts to help you understand the situation better. No one thinks "Gee, when I respond to a situation based on new facts I am applying this complicated statistical equation". We do it naturally. If someone calls the police saying they saw a wanted suspect in a specific location and say they provide decent surveillance footage of the suspect the police aren't going to plug in some mathematical equation to figure out they should go check it out to determine the suspect is probably around this area. They will naturally go check it out.

    • (Score: 2) by mindriot on Friday March 14 2014, @09:15PM

      by mindriot (928) on Friday March 14 2014, @09:15PM (#16631)

      Yep. What's more, saying that Bayes' Rule is what it's all about is a gigantic overstatement. The main clue to take away here is that it pays off to use a formal probabilistic model to fuse the available data sources. Bayesian inference plays a part in this, no doubt. But it doesn't build the probabilistic models for you. You need to take into account

      • A high-quality model describing the plane's motions and their associated uncertainties.
      • A good idea of the last reliably known location of the plane and, again, the uncertainty of that estimate.
      • Similar models for all additional data sources you may have had (e.g. triangulation of received radio signals, verbal messages from the plane's crew, weather reports and estimates of wind speeds, etc.). Again, with good estimates of the uncertainties of each individual data source.

      Then, and only then, can you use probability theory, including Bayes' Theorem, to infer a fused estimate of the probability distribution describing what you know about the plane's location. And use that estimate to guide your search.

      From my personal experience, getting a good grip on the true uncertainty of your data sources is usually the hardest part (anyone who's tried to parametrize a Kalman Filter for anything will know). If your model is overconfident, you'll have a small area to search, but risk wasting your time in the wrong area altogether. If the estimated uncertainty is too high, you'll have to search a needlessly large area, and the search will take forever.

      That said, I'm glad to hear that steps are taken to use a proper probabilistic approach, and that it can help reduce the time to discovery in practice. But TFA is pretty clueless.

      --
      soylent_uid=$(echo $slash_uid|cut -c1,3,5)
  • (Score: 3, Funny) by edIII on Friday March 14 2014, @05:08AM

    by edIII (791) on Friday March 14 2014, @05:08AM (#16183)

    Hadn't we already established it was aliens and that no black box could be found?

    Wait, sorry, I thought I was on Reddit for a moment...

    --
    Technically, lunchtime is at any moment. It's just a wave function.
    • (Score: 2) by frojack on Friday March 14 2014, @07:59AM

      by frojack (1554) on Friday March 14 2014, @07:59AM (#16213) Journal

      There is already more than one nut job putting up web pages to that effect.
      I would have thought it was a little too soon.

      --
      No, you are mistaken. I've always had this sig.
  • (Score: 4, Informative) by FakeBeldin on Friday March 14 2014, @09:10AM

    by FakeBeldin (3360) on Friday March 14 2014, @09:10AM (#16238) Journal

    P(A | B) = ( P(B | A) P(A) ) / P(B)

    In words:
    The chance of A happening, given that B is happening, is equal to:
    the chance that B is happening, given that A is happening, times the chance that A is happening, divided by the chance B is happening.

    (deliberately anon).

    • (Score: 4, Funny) by FakeBeldin on Friday March 14 2014, @09:17AM

      by FakeBeldin (3360) on Friday March 14 2014, @09:17AM (#16239) Journal

      Hey, who faked my name over that comment?!
      I distance myself from that, I would never stoop to any comment that useful ;-)

  • (Score: 4, Interesting) by Yog-Yogguth on Friday March 14 2014, @10:11AM

    by Yog-Yogguth (1862) Subscriber Badge on Friday March 14 2014, @10:11AM (#16257) Journal

    Seismologists (afaik) in China have detected a (presumably weak and/or noisy) seismic event in the Gulf of Thailand (that's the water the plane was flying over that lies between Malaysia, Thailand, Cambodia, and Viet Nam) which occurred roughly half an hour after the radar signal was lost. The waters are shallow (roughly 66 meters or 200 feet) and muddy and the plane might possibly have buried itself deep in the mud with minimal debris.

    Use your favorite search engine for links.

    --
    Bite harder Ouroboros, bite! tails.boum.org/ linux USB CD secure desktop IRC *crypt tor (not endorsements (XKeyScore))
  • (Score: 3, Funny) by JeanCroix on Friday March 14 2014, @01:29PM

    by JeanCroix (573) on Friday March 14 2014, @01:29PM (#16348)
    If we seed the Bayes theorem with the numbers 4, 8, 15, 16, 23, and 42, the exact location of the downed plane will be revealed. Until they turn the frozen donkey wheel.
  • (Score: 3, Interesting) by gottabeme on Friday March 14 2014, @02:45PM

    by gottabeme (1531) on Friday March 14 2014, @02:45PM (#16409)

    The Navy used Bayesian search to find a hydrogen bomb lost in the Atlantic back in 1966. It's described in the book Blind Man's Bluff [wikipedia.org].

  • (Score: 2) by Pslytely Psycho on Friday March 14 2014, @06:10PM

    by Pslytely Psycho (1218) on Friday March 14 2014, @06:10PM (#16543)

    ..... but may be successful anyway....

    IANAM (I am not a mathematician) but after reading the wiki, doesn't this just amount to 'deductive reasoning?'

    --
    Alex Jones lawyer inspires new TV series: CSI Moron Division.
  • (Score: 2) by umafuckitt on Friday March 14 2014, @06:28PM

    by umafuckitt (20) on Friday March 14 2014, @06:28PM (#16552)

    In the same way that Nate Silver worked independently to predict the results of the last US election using Bayes' Theorem, why don't statisticians work independently to predict the location of the missing MA flight? I'd really like to see their predictions evolve in real time as more information comes in.

  • (Score: 2) by MrGuy on Friday March 14 2014, @07:31PM

    by MrGuy (1007) on Friday March 14 2014, @07:31PM (#16590)

    Pasting multiple paragraphs out of TFA does not a "summary" make. In fact, it appears most of TFA's been copy pasted into the summary. If you can't summarize, don't try - just let us read the article.

    Also, only the first paragraph of TFS is indented in a way that makes it clear this is a quote and not original content by the editor. Maybe that's just how slashcode works, but it's not terribly awesome editorially to make this not distinct.

  • (Score: 0) by Anonymous Coward on Saturday March 15 2014, @01:20PM

    by Anonymous Coward on Saturday March 15 2014, @01:20PM (#16834)

    The plane stayed in the air for about 4 to 7 hours before crashing.

    The CVR only records the last two hours of cockpit voice.

    We'll never know how he took over the plane.