Time-Varying Environments and Noise

As we have seen, real-world problems are usually associated with large sample spaces and nonlinear, multimodal evaluation functions. In many cases, there are no closed-form mathematical procedures for generating solutions, and even when these techniques exist, they are often too computationally intensive to be practical. Their computational complexity increases at such a rapid rate that when we try to apply them to anything but trivial problems they are practically useless. In order to make them more useful, we accept simplified assumptions about the real world, perhaps mandating the application of a linear evaluation function, with linear constraints, integer values, or other common mathematical devices. We sacrifice the right answer just to get any answer at all.