For technical problems, one may struggle to define the specifications. When this happens, look at the behavioral design. Then one may find solutions from the design automation. Thankfully, evolution algorithms are a class of “soft computing” techniques which handles poor specifications. One will encounter direct application of the evolutionary algorithms via Neural networks, ReCaptcha, and / or Amazon Turk. All which have been applied to solve noisy pattern recognition.
Evolvable hardware techniques have important advantages over neural networks. Hardware is really fast and are more easily understood / implemented than neural networks with fast operations and solution tractability. For these purposes, evolvable hardware has been developed for academic, military and industrial applications. So let’s borrow from them instead of reinventing the wheel.
With appropriate automation, evolvable hardware has a large potential to autonomously adapt to the constantly changing risk environment. This is extremely useful for situations where (real-time approaches nano-seconds) over systems is improbable / impossible. It doesn’t hurt evolvable hardware is useful when harsh or unexpected data points, threats, vulnerabilities, risks, and other conditions are encountered.