Genetic Algorithms

This thread is meant to contain any question and/or material about GAs. Feel free to share or ask anything related.

There was someone in the last GA thread who said they had evolved a sorting algorithm. If they're still around, I'd quite like to see the algorithm that was produced, rather than just a gif of it working, if they have it.

Sorry for the ignorance, I wanted to build sonething like a neural net of sorts in Common Lisp for fun, but I have no idea where to start. What would the lainons recommend me to know?

Dropping this here, cause ay, tis a classic.

I am very interested in evolutionary computing. I am attracted the conceptual simplicity of it as well as the biologically-inspired design. I have made a simple regular expression generator. I want to extend it by keeping a database where new text strings can look for similar cases to use as the init population. Creative fitness functions and crossover/mutation schemes feel like you are working with the machines rather than directly programming them. Its like you have this source of random noise but influence it to converge to what you are looking for. >>34 Neural nets are a rather different approach than genetic algorithms. But I recently came across a paper where they used an evolutionary method to set the weights of the nodes in a ANN. It was really cool because they wanted to create a cellular automata but for a problem not entirely understood. So instead of writing the CA rules directly, the ANN's mappings of inputs to outputs implemented the CA rules. The rules were initially random, but evolved. I will attach the paper.

>>45 Than you so much for that, I'll take a careful look.

>>34 I'll redirect you to the new ML general: >>246 The two books in the OP are overkill if read cover to cover, but they cover a lot of stuff. And btw, yeah, like >>182 says, GA are not machine learning, though still artificial intelligence.

>>249 >GA are not machine learning, though still artificial intelligence On Wikipedia, GA are indicated as ML Can anyone any good book/doc about GA?

>>287 The thing about ML is that it's an algorithm you train to recognize patterns existing in a set of data; once the program has understood them, you can feed it new data and it'll tell you what it thinks about it. Genetic algorithms don't really do that, they're just a funny way of searching a state space for the optimum (at least afaik). Which is pretty much the definition of classic AI algorithms. If you want a reference, you can look at "Artificial Intelligence: A Modern Approach" by Stuart Russell and Peter Norvig, which is excellent and a classic. I'd attach it but either lainchan or my connection are being difficult, sorry; you can easily find it on libgen. It treats AI in general, including GA's. Mind the ML section is pretty barebones.

>>287 >>316 Sorry, to make it clear: ML is also search for an optimum in a state space (the hypothesis space of functions that "recognize patterns"); the difference is training over data, really.

I've tried to make a C implementation of the GA with Elitism but I must be retardely doing something wrong. What would it be the best place to ask help?

>>1592 Why not ask here? What is going wrong, and what have you tried? Sometimes just explaining your problem to others helps you see where the issue is.

>>1607 Oh, well, never mind. I paid a polite Ukranian developer who fixed my code and literally re-wrote it in a more elegant way.

>>316 >"Artificial Intelligence: A Modern Approach" by Stuart Russell and Peter Norvig ohai i'm reading that. yeah it's a good one, i looked through some real stinkers before finding it

Someone want to put up a simple example on github or propose a simple problem we could work on solving, as a fun exercise?

>>1629 Someone upload that soykaf.

>>317 What I would like to know is this: It looks like people use GA as a pre-processing step before conventional ML techniques. Why is this? Optimize inputs before further deduction?