"I'm no longer a skeptic, but still I can't resist the old skeptic urge to do a bit of debunking. After all, there are a lot of crackpots out there. There are people, for example, who believe that a superintelligent computer will arise in the next twenty years and then promptly either destroy humanity or cure death and set us free. There are people who believe that one of the best works of English literature is an unfinished Harry Potter fanfic by someone who can barely write a comprehensible English sentence. There are even people who believe the best thing you can do to help the poor and the starving is become a city stockbroker or Silicon Valley entrepreneur! And more often than not, the same people believe all these crazy things! The striking thing about these people is that they are no ordinary kooks — some of them actually identify as skeptics themselves, and all of them claim to be committed rationalists. Many are even full-time evangelists for "rationality", and can justify in sound and impressive detail why all their beliefs are correct. And they're not just backed up by the laws of logic and mathematics, but also by some of the finest minds and fattest wallets in Silicon Valley. Who are these people? They are the members of the Elect group who have received into their minds and hearts the glorious truth of something called Bayes' Theorem." - mkz

Bence olaya pek dogru bir acidan bakmamis (en azindan ML olaylari icin). Sadece rasyonalite ustunden gitmis - biraz istatistik felsefesi gibi olmus. Halbuki ML problemlerinde parametre uzerindeki belirsizligi cok iyi quantify ediyor Bayesian yontemler. Onlarsiz machine learning dusunemiyorum. :-) - D.

@D, karsi ciktigi sey su: "In general practice, there's no way to come up with meaningful figures for the right-hand side of the Bayes equation, and so Bayesians inevitably end up choosing values that happen to justify their existing beliefs." Yani, denklemin sag tarafini acik-secik degerlerle dolduramadigimiz zaman (mesela bir yazi-tura probleminin aksine), Bayesian yontemleri sonucun nasil olmasini istiyorsak ona gore kullaniyoruz (sosyal bilimlerde filan). Sen buna mi itiraz ediyorsun? - sizofroid‪‪‪‪‪‪‪

insanlar en basit ortamda bile beyezyen ogrenebilemiyor http://faculty.haas.berkeley.edu/hoteck... - hacıkenks

Kavga ciktiysa ben kahveden Andrew Gelman, Judea Pearl ve arkadaslari cagirayim, Stan kaynak kodunu da kapip gelsinler :-p (daha dun bir online dersin odevi icin bazi hastaliklara dair bir Bayes network ile ugrastim, tesadufun boylesi) - f(z)

Hayir sizofroid, ona itiraz etmiyorum - o dogru fakat pejoratif bir manada kullanilmasi yanlis. Abuse olamaz mi, olabilir; ama frequentist yontemler de 'kotuye kullanilamaz' diye bir sey yok ki. Her dandik bilimsel calisma Bayesian methodology'yi mi kullaniyor sanki? Frequentist ol, Bayesian ol (bir de boyle kesin ayrimlar da yok) her bilimsel problemi bastan bir model icinde ele aliyorsun bundan daha buyuk prior mu var? - Genel olarak rasyonalite kisminda ne donuyor bilmiyorum ama cikarim problemlerinde Bayesian yontemlerin alternatifi oldugunu sanmiyorum. - D.

Bir de benim o prior konusundaki fikrim soyle: Her calismada iyi veya kotu bir model var - fizikten tutun, machine learning'e kadar. Bir probability model yazdigin zaman ve bazi parametreler ustunde prior koydugun zaman, Bayesian framework bu konuda cok acik secik her seyi soylemis oluyor. Ne yaptigini direkt soyluyorsun - ne kadari objektif bilgi, ne kadari senin onyargin diye. Cogu non-Bayesian gorunen model, Bayesian contextte ifade edilebiliyor ve o zaman esasinda bir prior underlie ettigi gosterilebiliyor. Mesela image processing'de total variation'u minimize etmeye yarayan fluid dynamicsten alinma teknikler, Bayesian framework'de yazildiginda bir Markov random field ustunde smoothness prior'una karsilik dusuyor. Bayesci olunca 'zaten modele soylemissin abi' diyen adam, total variation minimization basliginda gorunce 'iyiymis' diyor. :-) - D.

Denk geldi: Isn't the distinction between Bayesians and everybody else right in the title--that one team uses Bayes's rule and everybody else doesn't? Bayes's rule has a one-line proof, and your typical stats/probability textbook covers it by page twenty. To not believe in Bayes's rule is equivalent to not believing basic standard probability theory. Models that don't explicitly use Bayes's rule at some point over the course of their typical description can frequently still be restated to make the underlying rule more apparent; explicit use of Bayes's rule is often computationally expensive, so Bayesian models are frequently reshaped to a version with no mention of Bayes's rule. - http://modelingwithdata.org/arch... - D.

D., bu arkadaşın asıl derdi o bahsettiğin 'rasyonalite' meselesiyle, ML olaylarına falan pek girmemesi ondan. O bahsettiği Lesswrong camiasına ve Yudkowsky'ye falan saydırmak için yazmış aslında yazıyı. Bayes'in ML'de falan kullanımı hakkında da senin söylediklerine yakın şeyler düşünüyorum ben de, ama tam netlediğim meseleler de değil bir yandan. Şunun altında da az linklediydik/lafladıydık: http://ff.im/19w8o8 - mkz

"One of Yudkowsky's constant refrains, appropriating language from Frank Herbert's Dune, is "Politics is the Mind-killer". Under this rallying cry, Lesswrong insiders attempt to purge discussions of any political opinions they disagree with. They strive to convince themselves and their followers that they are dealing in questions of pure, refined "rationality" with no political content. However, the version of "rationality" they preach is expressly politicised." - mkz

Bilmiyorum ne kadar haklı bir eleştiri bu, çok takip etmiyorum aslında (bi şekilde hep uyuz olduğum bi tarzı var sitenin (Lesswrong), ne zaman ilginç görünen bi şeyini açsam çok okuyamadan kapattım :), ama biraz öyle bir muhabbet var gibi gelmişti bana da. - mkz

COME WITH US IF YOU WANT TO LIVE - http://m.friendfeed-media.com/d831758... - mkz

Gorunmuyor ekledigin sey. - husdemir

PDF indirmesi lazımdı, denedim oldu şimdi ama şurdan da iniyor: https://pdf.yt/d... - mkz