Yeah, but Schmidhuber already had a paper on it in the 1890s.
Happy 70th birthday, k-nearest neighbor - the first true machine learning algorithm, capable of learning any function given enough data.
-
-
-
Can you share the link about this information? I know only about paper of Evelyn Fix and Joseph Hodges...
- Näytä vastaukset
Uusi keskustelu -
-
-
Shameless plug: turns out that k-NN is not universally Bayes-consistent, while a 1-NN based variant we proposed here (to appear Ann. Stat.) is https://www.e-publications.org/ims/submission/AOS/user/submissionFile/41585?confirm=c23bcd1a …
-
Also, Pelillo (2014, ref in in paper) traces the origins of 1-NN to Al-Hazen from 1000 years back.
Keskustelun loppu
Uusi keskustelu -
-
-
Tämä twiitti ei ole saatavilla.
-
I don't think so. Take the function on the real line that outputs 1 for all rational numbers and -1 for all irrational numbers. Good luck approximating that with a k-NN without memorizing the entire real line. But it should hold for reasonably regular functions.
- Näytä vastaukset
-
-
-
I read a good paper recently by
@Susan_Athey that helped me understand that random forests are like k-NN with an adaptive neighborhood size—serious big brain energy from Athey as usual. - Näytä vastaukset
Uusi keskustelu -
-
-
Very less known fact about this paper is the: co-author of this seminal paper is renowned information Theorist Late Thomas Cover (The author of the book elements of Information theory).
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
-
-
capable ... certainly not something that applies to your ethical comprehension.
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.