Efficiently Implementing Episodic Memory

Abstract

Abstract. Endowing an intelligent agent with an episodic memory affords it a multitude of cognitive capabilities. However, providing efficient storage and retrieval in a task-independent episodic memory presents considerable theoretical and practical challenges. We characterize the computational issues bounding an episodic memory. We explore whether even with intractable asymptotic growth, it is possible to develop efficient algorithms and data structures for episodic memory systems that are practical for real-world tasks. We present and evaluate formal and empirical results using Soar-EpMem: a task-independent integration of episodic memory with Soar 9, providing a baseline for graph-based, taskindependent episodic memory systems.

How Soar Agent Learns: Episodic Memory

“…Episodic memory (EpMem) in Soar is a mechanism that automatically captures, stores and temporally indexes agent state and supports a content-addressable agent interface to retrieve this autobiographical prior experience [4].…”

Scalable methods to integrate task knowledge with the Three-Weight Algorithm for hybrid cognitive processing via optimization

“…We see from this data that as the problem increases in size, the global reasoner maintains a much smaller problem graph and corresponding iteration time: even as the baseline TWA crosses 50 msec. per iteration, a commonly accepted threshold for reactivity 4 http://www.menneske.no/sudoku/eng in the cognitive-architecture community (Rosenbloom 2012;Derbinsky and Laird 2009;Derbinsky, Laird, and Smith 2010), our reasoner is able to easily maintain a real-time response rate for very large sudoku puzzles. Because our global reasoner modifies the problem graph, and thus the numeric optimizer must now search over a different problem, it was not obvious whether there would be an overall benefit in terms of time-to-solution.…”

Competence-Preserving Retention of Learned Knowledge in Soar's Working and Procedural Memories

“…Soar incorporates two long-term declarative memories, semantic and episodic . Semantic memory stores working-memory objects, independent of overall working-memory connectivity (Derbinsky, Laird, & Smith, 2010), and episodic memory incrementally encodes and temporally indexes snapshots of working memory, resulting in an autobiographical history of agent experience (Derbinsky & Laird, 2009). Agents retrieve knowledge from one of these memory systems by constructing a symbolic cue in working memory; the intended memory system then interprets the cue, searches its store for the best matching memory, and if it finds a match, reconstructs the associated knowledge in working memory.…”