Genetic Algorithms in Elixir Book Club!

Hello everyone! Excited that we’re doing this. Looking forward to learning from you all and the book!

I’d love to be a member

I have this book and this will be a really cool way of reading it. Looking forward to learn from you all!

Thanks for setting this up @AstonJ. I’m very excited about this book. I’ve started the book but only gone through the first couple of chapters. I’ll probably try to post every Wednesday with thoughts about the portion of the book I got through that week. I hope everyone else will do something similar. I know life is probably going to interfere at some point and I’ll fall behind, plus most, if not all, of the participants are going to bring more technical acumen to the discussion.

So far I’ll say I really enjoy the writing style. Concise without being dense. Precise without being obscure. I also like the repetition of the basic concepts between the first two chapters. Really helps to cement the ideas for me.

Looking forward to the coming weeks and hearing what everyone has to say about the book and the topic in general.

I’ll join in too!

I am starting chapter 1 now. Will update y’all tomorrow

I went over the first chapter. It comprehensively introduces the core steps of a genetic algorithm. The provided solution of the One-Max problem demonstrated that as the size of the problem scales up, it is necessary to consider further optimisations that I seek to find in the rest of the book. I tried to manually explore alternative values for the hyper-parameters but I was not overwhelmed by the performance results.
I looking forward to study the rest of the book.

P. S. An interesting question for me is whether genetic algorithms are suitable for matching two pieces of text, for example, in order to assess whether a patent claim extends beyond the original patent description.

Hi, I have a question, I tried running the one_max.exs script without mutation and it almost always gives me 1000 where the book mentions that it will be difficult for me to reach that. Adding mutation also gives me the same result (although it’s slower). So I was wondering if any of you could get the result matching what the book wrote?

UPDATE: I added an iteration counter and seems like (again based on the few tries I did), WITH MUTATION reaches the best result in ~50% less iteration count.

Indeed, also in my experiments the algorithm always reaches 1000 (even when disabling mutation).

I first wondered if it had to do with high machine specs but the book isn’t an old one and I think this has little to do with machine power and more to do with the way random number gets generated?

If I decrease the chromosome count but keep the bast value the same (thereby making it impossible to reach) then I get stuck at “current best: 10”. The algorithm doesn’t “stop” stop, but rather the data become stagnant and the algorithm tried forever. I tried multiple times but it is a solid 1000 each time