What problems are better solved by a genetic algorithm, and why?
In this project we tried to get some answers for a few theoretical questions: What kind of problems are better solved by a genetic algorithm, and why? What are the landscapes on which genetic algorithm outperforms other methods? Can genetic algorithms be used for discovery of optimal neural networks architectures? After discussing some experimental results obtained by the authors of (2), the paper also presents some experimental results using SANE, (Symbiotic, Adaptive Neuro-Evolution) (3), a Java package used to evolve (through genetic algorithms) a population of neurons to form a neural network for the sequential decision task. The primary paper discussed here is “Some Combinatorial Landscapes on which a Genetic Algorithm Outperforms Other Stochastic Iterative Methods” (2). The bibliography also includes several packages for neural networks and/or genetic algorithms (3,4,5,6), a general presentation of “Genetically Evolved Neural Networks” (7), and one of the best books about genetic
