Qualifiers for translating logical neural networks into natural language

From FMS-wiki
Jump to: navigation, search

Logical neural networks provide a tight integration between neural networks and logic: NNs are construed as logical formulas and vice versa. However, logical formulas are equipped with weights and biases (for learning) that are not directly interpretable. The goal is to examine whether and how the weights and biases can be interpreted in natural language, comparable to (but more complex than) qualifiers like "very" used in fuzzy logic.