For the last 20 years, my graduate students have explored explainable AI, multi-objective optimization, semi-supervised learning, configuration, planning, etc, etc. Can I summarize all the work? Code it up in under 1000 lines of code? Refactor it into a library of learning tools that can be mixed and matched as required?
Now that was interestinga . Underneath all these seemingly different applications is a small class library that can be mixed and matched in many ways. Here's an idea: before we leap to complex solutions (that are CPU hungry, that are hard to understand), perhaps we should try something simpler first. If the simpler methods work fine, then we can go home early. Else, we'd have a baseline approach which lets us appreciate the value of more complex solutions.