![runix cube sover runix cube sover](https://i.ytimg.com/vi/2nTT7FkghWw/maxresdefault.jpg)
And the alg insertion promise tough problems to solve. The point is to define an efficient fitness function for that task. ?solving all but some edges/corners? should be possible to implement, I will think about that. Actually a ?standard? solution (with ?normal? opposite 1x2x3 blocks) should not be too hard to program ? The drawback of his approach is the STM optimization ? but quite bad in HTM. Gilles' methods seems very interesting, but I doubt I could implement it easily in my program. Lars' general method is good but there is a need to know quite a lot o LL algs to be really efficient in terms of metricsįridrich is definitely not a good FM method ?I don't think any further discussion about this method is needed :) One of the best approaches to find a short solution (probably the best one ?) is Kociemba's alg, but as we all noticed there is no way one could figure out such a solution ? The question I am trying to answer is : ?what is the best mean to compute short, human-understandable solutions ? So now ? why am I telling you all this ? I thought it would be great to use your experience to improve the performance of the algorithms. The question was ? was it going to work ? I have no definitive answer, as no ?serious? validation was made and there are still lots of parameters I can tune in the genetic alg, but I fed it with #82 FMC scramble and here is I got a really cool 35 moves solution ( see details here) not bad for a first try in FMC challenge -) Get into 2-gen group, using edges orientation check and corners position check (see Sebastian's page) Try (1) for the 12 possible 2x2x3 rotations Try to solve the 2x2x3 in Back-Down in fewest moves as possible How could this be related to FMC ? Actually the progs I developed for my work could easily be translated to this problem, and I chose to adapt Charlie's and Sebastian's approach to solve the cube : To be short, genetic programming consists in defining a population of n individuals, then evaluate each individual with a fitness function, and use Darwin's rules of selection, mutation and reproduction to evolve to better individuals at each ?generation? and hopefully maximize the fitness function. In my academic work, which is related to medical physics, I have met a tough optimization problem and I am trying to solve it using genetic programming. Some recents FMC scrambles solved by the program are shown here : #81 ( 33 moves solution), #82 ( 35 moves solution), #83 ( 30 moves solution) and the last 2-gen challenge ( 23 moves solution) July, 5th, 2005 : Version 2.1 : Added one more feature to backup the scramble and the solution in a file. May 22nd, 2005 : I was finally able to write the "whole cube genetic solver 2.0" :-) Should solve any given or random scramble, most of the time between 30 and 40 moves for a completely scrambled cube. May 15th, 2005 : Here is the version 1.0 of the genetic solver ! At the moment, only 2x2x3 building is available, the rest will soon follow May 9th, 2005 : Some words about the beast (Forum message, Cyril Castella) May 7th, 2005 : Fewest moves Rubik?s cube solve using genetic algorithm ! (Forum message, Cyril Castella)Ĭomments and reactions (Messages from Yahoo forum)
![runix cube sover runix cube sover](https://i.ytimg.com/vi/1INglAm6rQI/maxresdefault.jpg)
I'll soon write a better description of the algorithms I am using, but the solver works quite fine now and it's the most important. Most of the subjects given here have been discussed on Yahoo fewest moves challenge forum, but I thought it would be a good thing to summarize every comments, discussions and ideas here. The aim of these pages is to describe the state of my project of designing a genetic rubik's cube solving program. Désolé, j'ai bien trop de choses à faire pour tout traduire en français, c'est pourquoi seule une version en anglais est disponible. Note aux francophones : ces quelques pages décrivent l'état d'avancement de mon projet de résolution du cube par un algorithme basé sur la programmation génétique.