3 ms·

How to become a Bayesian (2016)

50 points by noch 6y ago

fingerlocks 6y ago
I would really like to read the full paper for reason #3, but it’s paywalled. The whole article feels like a big teaser with the real content locked away.
- metasyn 6y ago
  scihub!
- sradman 6y ago
  https://zero.sci-hub.st/6400/5174e8f2acfdd03584eb6a754f9a3947/etz2017.pdf https://zero.sci-hub.st/6400/5174e8f2acfdd03584eb6a754f9a394...
- chalst 6y ago
  The TL;DR of the article provides two links to the preprint, and in fact there is a Springer version accessible without cost: https://link.springer.com/content/pdf/10.3758/s13423-017-1317-5.pdf https://link.springer.com/content/pdf/10.3758/s13423-017-131... Even less friction than Sci-Hub is the Unpaywall browser extension http://unpaywall.org/products/extension http://unpaywall.org/products/extension which finds links to unpaywalled content matching paywalled journal articles.
gdpr 6y ago
For anyone who wants to dig a bit deeper: PRML by Bishop [0]. An amazing work as a general introduction to machine learning and Bayesian in general. MacKay's book [1] is a bit more opinionated regarding Bayesian methods. Even if you do not become purely Bayesian, it advances your understanding of to approach data quiet a bit compared to the generic "oh data, lets throw a NN at it". [0] https://www.microsoft.com/en-us/research/people/cmbishop/prml-book/ https://www.microsoft.com/en-us/research/people/cmbishop/prm... [1] http://www.inference.org.uk/mackay/itila/book.html http://www.inference.org.uk/mackay/itila/book.html
woodandsteel 6y ago
About 40 years ago I was reading a business mathematics book and ran across a section that, as I recall, started out with a basic probabilities equation and derived the basic Bayesian equation in one step. Am I remembering this right? If so could someone post it?
- hgoury 6y ago
  By definition: p(A|B) = p(A,B) / p(B) (1) With the same definition, switching A and B roles, we have p(A, B) = p(B|A) p(A) (2) Plug (2) into (1) and you have Bayes rule.
  - woodandsteel 6y ago
    Thank you for the explanation.

50 points by noch 6y ago

I would really like to read the full paper for reason #3, but it’s paywalled. The whole article feels like a big teaser with the real content locked away.

metasyn 6y ago
scihub!
sradman 6y ago
https://zero.sci-hub.st/6400/5174e8f2acfdd03584eb6a754f9a3947/etz2017.pdf https://zero.sci-hub.st/6400/5174e8f2acfdd03584eb6a754f9a394...
chalst 6y ago
The TL;DR of the article provides two links to the preprint, and in fact there is a Springer version accessible without cost: https://link.springer.com/content/pdf/10.3758/s13423-017-1317-5.pdf https://link.springer.com/content/pdf/10.3758/s13423-017-131... Even less friction than Sci-Hub is the Unpaywall browser extension http://unpaywall.org/products/extension http://unpaywall.org/products/extension which finds links to unpaywalled content matching paywalled journal articles.

gdpr 6y ago

For anyone who wants to dig a bit deeper: PRML by Bishop [0]. An amazing work as a general introduction to machine learning and Bayesian in general. MacKay's book [1] is a bit more opinionated regarding Bayesian methods. Even if you do not become purely Bayesian, it advances your understanding of to approach data quiet a bit compared to the generic "oh data, lets throw a NN at it". [0] https://www.microsoft.com/en-us/research/people/cmbishop/prml-book/ https://www.microsoft.com/en-us/research/people/cmbishop/prm... [1] http://www.inference.org.uk/mackay/itila/book.html http://www.inference.org.uk/mackay/itila/book.html

woodandsteel 6y ago

About 40 years ago I was reading a business mathematics book and ran across a section that, as I recall, started out with a basic probabilities equation and derived the basic Bayesian equation in one step. Am I remembering this right? If so could someone post it?

hgoury 6y ago
By definition: p(A|B) = p(A,B) / p(B) (1) With the same definition, switching A and B roles, we have p(A, B) = p(B|A) p(A) (2) Plug (2) into (1) and you have Bayes rule.
- woodandsteel 6y ago
  Thank you for the explanation.