LTNtorch’s documentation

Welcome to the LTNtorch’s documentation!

LTNtorch is a fully tested and well documented PyTorch implementation of Logic Tensor Networks (LTNs), a Neural-Symbolic approach which allows learning neural networks using the satisfaction of a First-Order Logic (FOL) knowledge base as an objective.

The documentation is organized as follows:

• Notes: contains some information that may be useful for those unfamiliar with the LTNtorch framework;

• LTNtorch’s modules:

  • ltn.core, which contains the definition of constants, variables, predicates, functions, connectives, and quantifiers;

  • ltn.fuzzy_ops, which contains the definition of some of the most common fuzzy semantics (connective operators and aggregators).

Notes

• Grounding in Logic Tensor Networks
• Introduction to Learning in Logic Tensor Networks
• LTN objects
• LTN broadcasting
• Quantification in Logic Tensor Networks
• Stable Fuzzy Semantics

Modules

• ltn.core
• ltn.fuzzy_ops