Description

Weather plays a significant role in terms of life, property, agriculture and industry. Neural networks are capable of predicting the non-linear behavior of weather without the physics being explicitly explored. The most common method to train neural networks is through gradient decent based back propagation algorithm. But back propagation algorithm suffers from several disadvantages like local minima problem, slow training, and scaling problem. So the ways to solve these problems by hybridizing it with genetic algorithms. The hybrid technique can learn efficiently by combining the strengths of genetic algorithm with back propagation algorithm . The hybrid neural networks are more qualified if only the requirement of a global searching is considered. It is good at global search i.e. not in one direction and it works with a population of points instead of a single point. Also it blends the merits of both deterministic algorithm BP and stochastic optimizing algorithm GA.