Crossover - Writing your own Genetic Algorithm Part 3 

Thanks for the kind words, I don't think anyone has ever complimented my voice before haha :)

You’re welcome! Thanks for watching

Thanks, I appreciate that & am glad you found it helpful

Hey thanks for the feedback! OOP can be a really tricky thing to get your head around to start with, but once you do, it feels as natural as breathing :) I'm glad you're enjoying the series

@@TutorialsWithGary Finished, all is working well. I learnt a great deal so thanks again. My main takeaways from this: 1. I should have been more modular in my approach to designing classes. I ended up with 4 or 5 classes which did everything. I think a better approach would have been to use more classes with fewer responsibilities each. 2. I didn't exploit inheritance, and I would do if I was to do this again. For example, I had an 'Individual' Class and a parent class, which had identical properties but different methods. I should have made the parent class a child of the individual class. 3. I ended up with lots of duplicated individuals in the population after a few generations which led to finding local minima rather than the global minimjm. My approach to overcoming this was inefficient, with nested loops replacing duplicates with a new individual. I guess this could be considered as a mutation, but it slowed down my solution a lot. 4. I built this specifically to solve a world with 10 cities. I had to do a good deal of redesigning to abstract this to solve n cities, which is due to poor planning at the start. again, thanks for this video series.

Wow that sounds like an incredible outcome, I love the drive. A lot of your takeaways are quite important, and are guaranteed to improve and just become more natural over time. The code I have been writing in this series could definitely be technically better - but it can make things less clear for people less familiar with the language following along. In my next video I'm going through the implementation of multi-objective, which I took as an opportunity to refactor a bunch of the code, and add a test project into the mix. If you're interested in looking over some improved quality code, I'd recommend giving it a look! github.com/Gary-The-Cat/MultiObjective

That’s an interesting approach. In terms of what is ‘better’ - it’s hard to say. But it’s good to think about it in terms of what the point of our crossover operator is - to take characteristics of our parents that make them genetically good, and pass those on to the children. For the implementation I talked through, randomly splitting gives the chance to take two separate good parts from both parents. How do you imagine your implement would do? I think both implementations would work - hard to say what is better without some testing.

Hey, how are your solutions encoded & what are they representing ? The best crossover operator can be hard to choose.

@@TutorialsWithGary Here two randm solutions are for example A,B. By taking these two random solutions are combining with crossover to generate random solution C in genetic algorithm with TSP. I need help how can I implement. Please suggest me and can you give me email address