A non connectionist way for ML (pipe dream)
Uses a simple logic engine, but rules are associated with confidence and rearranged based on the feedback loop from processed data. Reflects more of my ignorance of topics and languages than anything bright.
Inspired from prolog, NLP and everything else that I hardly understand.
dimsum is dimensional summarizer, it looks at the problem from an expert's dimension and captures his summary of thumbrules about the solution. These rules are then rearranged based on data feedback. New rules can also be inferred based on new data, but the main goal is that all rules must be readable by a human. As an example look at the app processing rules in examples folder.