Key-value memory in the brain

2025-01-06

AI Summary

Existing memory theories fail to explicitly distinguish between stored information (values) and retrieval cues (keys), limiting mechanistic understanding of memory. This paper proposes and systematically develops a computational framework for key-value memory in the brain, transcending traditional similarity-based retrieval models to jointly optimize storage fidelity and retrieval discriminability. Methodologically, we integrate computational neuroscience modeling, key-value attention mechanism analysis, cognitive psychological experimentation, and biologically plausible inference. We first demonstrate that key-value mechanisms confer fundamental advantages over classical associative memory in functional decoupling and computational efficiency. Second, we unify explanations of long-standing memory paradoxes—including context-dependent forgetting and rapid pattern separation. Third, we formulate a testable neurobiological hypothesis specifying how key-value operations may be implemented within the hippocampal–neocortical circuitry.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Humans and AI: Learning Human Values and Preferences
• Search and Optimization: Metareasoning and Metaheuristics

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Psychology-informed user models and recommender systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Classical models of memory in psychology and neuroscience rely on similarity-based retrieval of stored patterns, where similarity is a function of retrieval cues and the stored patterns. While parsimonious, these models do not allow distinct representations for storage and retrieval, despite their distinct computational demands. Key-value memory systems, in contrast, distinguish representations used for storage (values) and those used for retrieval (keys). This allows key-value memory systems to optimize simultaneously for fidelity in storage and discriminability in retrieval. We review the computational foundations of key-value memory, its role in modern machine learning systems, related ideas from psychology and neuroscience, applications to a number of empirical puzzles, and possible biological implementations.

Problem

• Key-Value Memory
• Brain Memory Mechanisms
• Biological Implementation

Innovation

• key-value memory system
• storage and retrieval differentiation
• biological implementations in machine learning