Neural networks that convert visual input features into intermediate representations via weights and activation functions suffer from a lack of direct observability and testability, which limits their interpretability despite the use of logical expressions in the reasoning process.

Authors

• Person: Not available Organization: arXiv
  Neuro-Symbolic AI: Explainability, Challenges, and Future Trends

Sources

• Neuro-Symbolic AI: Explainability, Challenges, and Future Trends arxiv.org arXiv via serper

Referenced by nodes (1)

• artificial neural networks concept