After a certain point, even an extremely complex solution ceases to improve. And then it indicates, we have to improve the design of the solution. We go back to the drawing board and put our heads together to improve the capabilities. With more options available, we can iterate and test multiple solutions. And then based on the business problem at hand, the best solution will be chosen and implemented. We follow the same principle in deep learning architectures. We work on a network architecture, improve it, and make it more robust, accurate, and efficient. The selection of the neural network architectures is based on the testing done of various architectures.