From numerical proportions to analogical proportions between probabilities

This academic paper extends analogical proportion theory from numerical and vector-based representations to probabilistic settings, investigating whether probability distributions associated with analogically proportional profiles maintain proportional relationships. The research bridges formal logic with statistical inference, potentially enabling more sophisticated classification methods that operate on probabilistic data.

This theoretical paper addresses a fundamental question in formal logic and machine learning: whether analogical reasoning patterns hold across different mathematical representations. The authors build on established work showing that when four profiles form analogical proportions at the component level, their class labels tend to maintain that proportional structure. They now extend this framework to probability distributions, examining whether profiles whose frequency distributions describe discrete attributes can themselves form analogical proportions while preserving normalization constraints.

The research stems from decades of work in analogical reasoning and its application to classification systems. Analogical proportions—formalized as "a is to b as c is to d"—offer an intuitive yet mathematically rigorous approach to pattern recognition that differs fundamentally from traditional statistical methods. Previous studies demonstrated this approach's effectiveness for vector-based data, motivating exploration into more complex mathematical structures.

The significance lies in potential applications to probabilistic machine learning and uncertainty quantification. If analogical proportions reliably transfer to probability distributions, this could enable new classification architectures that reason about uncertain data while maintaining interpretability. The work bridges symbolic AI reasoning with probabilistic inference, two traditionally separate domains.

The authors compare definitions based on arithmetic and geometric proportions, evaluating which approaches best preserve distributional properties. Their experimental investigation determines whether analogically proportional profiles generate analogically proportional probability distributions—a crucial property for extending the theoretical framework. Success here could influence how researchers approach classification under uncertainty and inspire hybrid systems combining logical reasoning with statistical methods.

• →Research extends analogical proportion theory from numerical vectors to probabilistic distributions with normalization constraints.
• →Previous work showed analogical proportions in profiles predict proportional class relationships; this paper tests if probability distributions behave similarly.
• →The study compares arithmetic and geometric proportion definitions to determine which best preserves distributional analogical properties.
• →Successful results could enable new classification methods that reason probabilistically while maintaining interpretable analogical relationships.
• →The work bridges symbolic AI logic with statistical inference, combining two traditionally separate computational approaches.