IID Relaxation by Logical Expressivity: A Research Agenda for Fitting Logics to Neurosymbolic Requirements

本文提出，神经符号背景知识及其逻辑所需的表达能力可以打破机器学习对数据独立性和相同分布性的假设。我们建议在适合不同用例要求的逻辑层次结构中分析IID松弛。我们讨论了利用已知数据依赖性和分布约束条件的神经符号用例的好处，并认为为此所需的表达能力对底层机器学习程序的设计具有影响。这开启了一个新的研究议程，涉及神经符号背景知识和其逻辑所需的表达能力的一般问题。

Neurosymbolic background knowledge and its logic's expressivity can challenge Machine Learning assumptions about data independence and identical distribution. The paper aims to analyze IID relaxation in a hierarchy of logics that fit different use case requirements.

The paper proposes leveraging known data dependencies and distribution constraints for Neurosymbolic use cases, which requires the expressivity of its logic. This opens a new research agenda with general questions about Neurosymbolic background knowledge and the expressivity required of its logic.

The paper discusses the benefits of using known data dependencies and distribution constraints for Neurosymbolic use cases. It also argues that the expressivity required for this knowledge has implications for the design of underlying ML routines. The experiments are not discussed in detail, and there is no mention of open-source code. The paper suggests a new research agenda for Neurosymbolic background knowledge and its logic's expressivity.

Related work includes recent research in the Neurosymbolic field, such as 'Neuro-Symbolic Concept Learner: Interpreting Scenes, Words, and Sentences From Natural Supervision' and 'Neuro-Symbolic VQA: Exploring Methods for Fusing Logic and Deep Learning Models'.