I will join in as much as I can, sounds really interesting.
rvirding
Creator of Erlang
stevensonmt
That’s a good catch on Chapter 2, but the way individual problems gets addressed is refactored in Chapter 3 so that the order of parameters is not an issue when Genetic.run is called.
You’ve hit on something that I’ll mention when I put together my thoughts more cleanly, but coming from a medical/biology background the metaphor is slightly skewed. As you’ve pointed out a genotype refers to the makeup of an organism, not of an individual chromosome. And populations are collections of organisms, not collections of chromosomes. So really the Russian doll terminology should be population -> organism -> chromosome -> gene -> allele. It’s not a substantive issue and probably not something specific to this book, but it requires a little shift in my head every time.
2
stevensonmt
I’m curious if you’re aware of any movement to use Erlang in this type of problem. Given how @AstonJ has categorized this thread in the Nx Forum I’m assuming the book is headed toward using that library extensively. If not, are there any particular features of Elixir that make Nx possible where it would not be in Erlang?
3
AstonJ
I haven’t read the book myself, but I put the thread here as generally I feel the Nx section is a good fit for all ML/A.I type threads (would you agree?)
1
code-shoily
Now that you folks bring it, I remembered a question I asked myself when I last picked this book.
How much of this book will change due to the arrival of the Nx mix of tools? Beyond just converting all
EnumintoNx?
This might as well be a silly question as I don’t have much exposure to ML, but in general I see Python folks bring Numpy/Pandas the moment they smell a puff of data. Does interfacing with any C++ backend for genetic algorithm via Nx feel like a feasible direction to aim at?
Again, I might be asking this entirely prematurely, some stuff can perfectly be done through standard lib.
5
stevensonmt
Certainly seems reasonable. And if nothing else explorers can maybe add to the book by experimenting with Nx if the authors don’t go that route. (I haven’t gotten past chapter 4 and don’t have access at work right now to check if the book already uses Nx or not.)
3
polvalente
Nx Core Team
The book is actually what spawned the idea for Nx. Unless it was updated for a second edition (which I’m unaware of), I don’t think it has Nx code.
I’d be more than happy to help with translating things to Nx, though. I probably won’t have enough time to actually participate in the book club, but I can help with code 
6
stevensonmt
m4hi2
Would love to have the livebook repository along with visualizations. It’ll help to understand the algorithms better.
5
rvirding
Creator of Erlang
I honestly don’t know. I have mainly so far done it for personal interest.
From what I can see and through some simple tests Nx seems to work well with Erlang. I have at least testade the simple examples given in the “Nx for absolute beginners” and “Up and running Nx”. I made a very simple Erlang module called nx which just passes all the calls through to Nx. As there a re no macros in Nx there are no problems calling it from Erlang.
The thing to remember is that Elixir runs on top of Erlang/BEAM so everything it does goes through them so it can all be done directly from Erlang. The question is more what the interface looks like. As soon as it includes Elixir macros then it becomes more complex to handle. And the module names don’t help. 
I must also admit I have another interest. I have been testing from LFE 
This is about the same as for Erlang but if I ever get to the stage of needing macros then LFE is much easier than Erlang. And I like the parentheses.
We shall see how it goes with the book.
3