This reminds me of the Sudoku puzzle solver I made one time for a minor school project using the so-called genetic algorithm. It turns out that it is really hard (maybe impossible, at least in my experience) to find 'fitness rules', that allow for a movement toward the optimal solution, this being the solution of a sudoku puzzle, using a population of filled-in sudoku grids and fitness rules, without applying any normal logic to deduce numbers based on the given numbers. The algorithm would move most of the population toward one of the 'lower peaks', satisfying the fitness rules best it could. However because one wrong placement leads to another, the end result would most often look nothing like the optimal solution - no way to get from the 'lower peak' to the optimal peak. Even with a lot of the numbers given, there would be a chain of bad numbers influencing each other's placement just sitting there, maybe mutating here and there but always moving back toward that suboptimal solution. My conclusion for that project was of course that the genetic algorithm is not a good way to find optimal solutions, but where a decent solution suffices (famous example being the traveling salesman problem) it shows relatively good results. In the IT world that is, not in real life ofc.
The majority of people on UA-cam (and in America especially) do not have much scientific training and this is pretty far over their heads. My observation is only a very small percentage of the population is interested in science. Our culture is heavily entertainment and comfort focused. And happy to believe "experts" without checking out the data or study
This reminds me of the Sudoku puzzle solver I made one time for a minor school project using the so-called genetic algorithm.
It turns out that it is really hard (maybe impossible, at least in my experience) to find 'fitness rules', that allow for a movement toward the optimal solution, this being the solution of a sudoku puzzle, using a population of filled-in sudoku grids and fitness rules, without applying any normal logic to deduce numbers based on the given numbers.
The algorithm would move most of the population toward one of the 'lower peaks', satisfying the fitness rules best it could. However because one wrong placement leads to another, the end result would most often look nothing like the optimal solution - no way to get from the 'lower peak' to the optimal peak. Even with a lot of the numbers given, there would be a chain of bad numbers influencing each other's placement just sitting there, maybe mutating here and there but always moving back toward that suboptimal solution.
My conclusion for that project was of course that the genetic algorithm is not a good way to find optimal solutions, but where a decent solution suffices (famous example being the traveling salesman problem) it shows relatively good results. In the IT world that is, not in real life ofc.
Yes, thank you. I'm amazed that this channel doesn't have more followers and comments. I always benefit greatly. 🙏☦️🏅
I'm here, and I agree.
@@georgemonnatjr.172 ha, I happen to be a George Junior as well. Cheers
The majority of people on UA-cam (and in America especially) do not have much scientific training and this is pretty far over their heads. My observation is only a very small percentage of the population is interested in science. Our culture is heavily entertainment and comfort focused. And happy to believe "experts" without checking out the data or study
Thank you for promoting Intelligence!
Blessings
Excellent post. Visuals would be great help here, but difficult to produce, especially for lay person/high school level