Neural-symbolic computing is a research are that aims at integrating the ability of neural networks to learn from the environment and the ability to reason using symbolic representation of knowledge. Please see my relevant research papers:

• Fuzzy Non-Horn Knowledge Bases: Calculi, Models, Inference
• Calculi and Models for Non-Horn Knowledge Bases Containing Neural and Evaluable Predicates
• Symbolic Inference for Non-Horn Knowledge Bases With Fuzzy Predicates
• Heuristic Backward Chaining Based on Predicate Tensorization
• A Best-First Backward-Chaining Search Strategy Based on Learned Predicate Representations
• Hierarchical Rules for Knowledge Representation and Learning

Need to relax? Try brain teasers. I would recommend those marked 'cool'.