aka "Bayesian Methods for Hackers": An introduction to Bayesian methods + probabilistic programming with a computation/understanding-first, mathematics-second point of view. All in pure Python ;)
First of all, thank for writing a very great book, I'm really enjoying it :smile:
Currently in the TFP version of this book, models are "manually" built. I took a look at TFP tutorial here, found out that models could be intuitively built with JointDistribution*{AutoBatched}. For some simple models, this has some advantage of its own:
intuitive: can be read seamlessly between code and visualization graph, "fast prototype Bayesian model" to quote their words
handles sample() and log_prob() together, you don't need to write different functions
So I suggest you could add this new syntax to your book to keep it up-to-date with new version of TFP
Hello,
First of all, thank for writing a very great book, I'm really enjoying it :smile:
Currently in the TFP version of this book, models are "manually" built. I took a look at TFP tutorial here, found out that models could be intuitively built with
JointDistribution*{AutoBatched}
. For some simple models, this has some advantage of its own:sample()
andlog_prob()
together, you don't need to write different functionsSo I suggest you could add this new syntax to your book to keep it up-to-date with new version of TFP
Thank you