On the Right Path: A Modal Logic for Supervised Learning

Abstract

Formal learning theory formalizes the process of inferring a general result from examples, as in the case of inferring grammars from sentences when learning a language. Although empirical evidence suggests that children can learn a language without responding to the correction of linguistic mistakes, the importance of Teacher in many other paradigms is significant. Instead of focusing only on learner(s), this work develops a general framework-the supervised learning game (SLG)-to investigate the interaction between Teacher and Learner. In particular, our proposal highlights several interesting features of the agents: on the one hand, Learner may make mistakes in the learning process, and she may also ignore the potential relation between different hypotheses; on the other hand, Teacher is able to correct Learner's mistakes, eliminate potential mistakes and point out the facts ignored by Learner. To reason about strategies in this game, we develop a modal logic of supervised learning (SLL). Broadly, this work takes a small step towards studying the interaction between graph games, logics and formal learning theory.

What Do Hebbian Learners Learn? Reduction Axioms for Iterated Hebbian Learning

“…Our approach to modeling iterated Hebbian learning takes inspiration from dynamic epistemic and doxastic logics (DELs) (Van Ditmarsch, van Der Hoek, and Kooi 2007;Van Benthem 2011). Two recent papers, (Baltag, Li, and Pedersen 2019; present DELs that model an agent's learning. But it is unclear how these learning policies might relate to specific neural implementations of learning such as Hebbian update and backpropagation.…”

Bisimulations for Knowing How Logics

“…There is also the drastically different knowing why , concerning the reasons/arguments/evidence/justifications the agent has for claiming that a certain proposition is the case [6, 64]. There are also several forms of knowing what , discussing variables, their possible values, their possible dependencies, and knowledge about the latter two (e.g., [8, 26, 57]). In [61] Wang presents a survey of logical approaches to these forms of knowledge.…”