Machine learning with rules instead of data

I have a AI problem in which I don’t have data, but relatively big, complex and fuzzy set of rules given by domain experts. Natural approach would be to implement these rules by rule tools or languages like JESS. However, I am not familiar of those, but I AM familiar of ML and therefore would like to use ML to implement the rules. Basically, my idea is the following: 1) generate artificial data using the rules 2) create ML model using the generated, artificial data set I tried to search from web some papers or other info on this approach, but couldn’t find any. Has anybody came across this kind of idea? /// “When you have a hammer, every problem looks like a nail”

Link to Full Article: Machine learning with rules instead of data