On Bringing Case-Based Reasoning Methodology to Deep Learning

摘要

The case-based reasoning community is successfully pursuing multiple approaches for applying deep learning methods to advance case-based reasoning. This "Challenges and Promises" paper argues for a complementary endeavor: pursuing ways that the case-based reasoning methodology can advance deep learning. Starting from challenges in deep learning and proposed neural-symbolic integrations based on specific technologies, it proposes studying how CBR ideas can inform choices of components for a new reasoning pipeline.