Composing Neural Learning and Symbolic Reasoning with an Application to Visual Discrimination
This addresses the need for interpretable and verifiable AI systems in visual tasks, though it appears incremental as it combines existing neural and symbolic methods.
The paper tackles the problem of creating interpretable discriminators for visual classification tasks by introducing Visual Discrimination Puzzles (VDP), which require classifying images according to logical specifications. The authors propose a neurosymbolic framework combining neural networks with symbolic reasoning, showing it performs favorably compared to purely neural approaches on VDP datasets.
We consider the problem of combining machine learning models to perform higher-level cognitive tasks with clear specifications. We propose the novel problem of Visual Discrimination Puzzles (VDP) that requires finding interpretable discriminators that classify images according to a logical specification. Humans can solve these puzzles with ease and they give robust, verifiable, and interpretable discriminators as answers. We propose a compositional neurosymbolic framework that combines a neural network to detect objects and relationships with a symbolic learner that finds interpretable discriminators. We create large classes of VDP datasets involving natural and artificial images and show that our neurosymbolic framework performs favorably compared to several purely neural approaches.