mahdighodsi
commented
10 years ago - For Initializing the population since right now we are purely choose random populations based on my research we can improve our final result by using other induced initializing method. However, using purely induced would add a lot of complexity to our code. Therefore, the most ideal solution for initializing the population based on my research is a mixture of random and induced method. This method would generate the initial city randomly and continue listing the nodes from the shortest distance to the farthest node by using the distance matrix that we already implemented. This method has proven to generate a better overall solution. -For improving the algorithm by introducing crossover, the simplest methods that can replace some of our mutation methods would be in order of complexity single point, double point and cut and slice crossovers. In addition, one of the most effective cross over is know as Edge Recombination which I have found a sample code that we can test in the following link: https://github.com/estsauver/GAOT/blob/master/enhancederXover.m tutorial: http://www.rubicite.com/Tutorials/GeneticAlgorithms/CrossoverOperators/EdgeRecombinationCrossoverOperator.aspx
rohams
-Investigating a new method for initializing the population (currently is completely random) to avoid immature convergence. -Improving the algorithm by introducing crossover or new permutation techniques. (I will also contribute to this) -Developing a fitness value that considers capacity constraint in addition to off-routing.