Re:Bayesian Optimization Book
by armcat on Monday November 15, 2021 @07:20AM
(#29226510)
Looks really promising, will give it a read through!
For those looking for an easier entry into Bayesian analysis, as a related topic and a starting place before embarking on Bayesian optimization, I would highly recommend "Statistical Rethinking" by Richard McElreath: https://xcelab.net/rm/statistical-rethinking/. Why I really like Richard's book is that it bypasses lot of the heavy mathematical/integral work, and goes straight into sampling - from my experience, you generally can't do integrals by hand, you describe your model in terms of a hierarchy of probability distributions and let an MCMC sampler take care of the rest. Richard's book touches upon causality (important and often overlooked topic in ML!), and you can follow his course online: https://github.com/rmcelreath/stat_rethinking_2022
EDIT: made my recommendation more explicit as a related topic.
Re:Bayesian Optimization Book
by abhgh on Monday November 15, 2021 @07:50AM
(#29226706)
Maybe you meant your recommended book as related reading, but the linked text is for Bayesian optimization, which is different from Bayesian analysis. Sure the optimization uses some Bayesian ideas but the majority of the challenges are different.
For those looking for an easier entry into Bayesian analysis, as a related topic and a starting place before embarking on Bayesian optimization, I would highly recommend "Statistical Rethinking" by Richard McElreath: https://xcelab.net/rm/statistical-rethinking/. Why I really like Richard's book is that it bypasses lot of the heavy mathematical/integral work, and goes straight into sampling - from my experience, you generally can't do integrals by hand, you describe your model in terms of a hierarchy of probability distributions and let an MCMC sampler take care of the rest. Richard's book touches upon causality (important and often overlooked topic in ML!), and you can follow his course online: https://github.com/rmcelreath/stat_rethinking_2022
EDIT: made my recommendation more explicit as a related topic.