An Investigation for Intelligence and Classical Optimization for NLP

This work is concerned with the improvements of both intelligence optimization technique (such as genetic algorithm) and the classical optimization technique (such as conjugate gradient) for non-linear unconstrained and constrained problems. This work contains the hybrid intelligence and classical optimization techniques. In the first part of this work deals with unconstrained problems and other part deals with constrained problems. Firstly, we have the classical conjugate gradient methods with scaling and restarting techniques based on the crossover of the genetic algorithm. Secondly, the modified genetic algorithm is used to solve constrained problems depending on the transformation of function and variables values. The integer-programming problem with genetic algorithm is also investigated. Finally, in this work a new conjugate gradient neural network and hybrid Fletcher Reeves type method have been investigated. Our improvement on conjugate gradient methods and genetic algorithms show that its promising when compared with other standard algorithms to solve non-linear optimization problems.