A Biologically Interpretable Cognitive Architecture for Online Structuring of Episodic Memories into Cognitive Maps
E. A. Dzhivelikian, A. I. Panov
TL;DR
This work tackles how cognitive maps and hippocampal-like structures can emerge from online, biologically plausible learning rules. It proposes a two-level memory architecture where the first level perfectly stores episodic sequences via a deterministic HMM, and a second level merges first-level states into clusters using Successor Features (SF) similarity, enabling generalization while preserving information with a discount factor $\gamma \in (0,1)$. In a partially observable grid-world, SF-guided merges improve next-observation prediction and better reflect the environment's transition structure, while reducing the second-level state space. Overall, the approach advances biologically grounded cognitive mapping in AI by integrating local Hebbian-like learning with SF-based memory structuring and a two-tier memory hierarchy.
Abstract
Cognitive maps provide a powerful framework for understanding spatial and abstract reasoning in biological and artificial agents. While recent computational models link cognitive maps to hippocampal-entorhinal mechanisms, they often rely on global optimization rules (e.g., backpropagation) that lack biological plausibility. In this work, we propose a novel cognitive architecture for structuring episodic memories into cognitive maps using local, Hebbian-like learning rules, compatible with neural substrate constraints. Our model integrates the Successor Features framework with episodic memories, enabling incremental, online learning through agent-environment interaction. We demonstrate its efficacy in a partially observable grid-world, where the architecture autonomously organizes memories into structured representations without centralized optimization. This work bridges computational neuroscience and AI, offering a biologically grounded approach to cognitive map formation in artificial adaptive agents.