A cellular Automata(CA) is a system of finite automata that provide a discrete computational model to under the complex behaviour of images (computational and real).This article provide a survey on how researchers can train the Cellular Automata to learn best rules to achieve optimal solution in a large search space. Many researchers have provided different-different methods (SFFS, with GA, 3-State Representation ....) to train .CA trained with GA(Genetic Algorithm) capable to perform various difficult tasks. This survey also introduces the different fields where CA used with modification techniques (B-Rules CA, 2-Cycle CA).