# Associative Memory **Domain:** Neural Networks / Memory / Cognitive Modeling **Doc Type:** Canonical Concept Node **Maturity:** Foundational ## Definition **Associative memory** stores relationships in a distributed structure so that a complete pattern can be reconstructed from a partial, noisy or related cue. ## 1972 Convergence In 1972, [[wiki/James A. Anderson|James A. Anderson]], [[wiki/Teuvo Kohonen|Teuvo Kohonen]], [[wiki/Kaoru Nakano|Kaoru Nakano]] and [[wiki/Shun-Ichi Amari|Shun-Ichi Amari]] published related associative-memory architectures across the United States, Finland and Japan. ## Technical Context The memory is not retrieved from one address. It is regenerated through the interaction of weighted relations, making graceful degradation and cue-based completion possible. ## See Also [[wiki/Distributed Representation|Distributed Representation]], [[wiki/Hopfield Network|Hopfield Network]], [[wiki/Recurrent Neural Network|Recurrent Neural Network]] ## Simple Reminders, Quotations, and Thoughts > “Computer memory is much better than human memory because early computer scientists discovered a trick that evolution never did: organizing information by assigning every memory to a sort of master map, in which each bit of information that is to be stored is assigned a specific, uniquely identifiable location in the computer's memory vaults. Human beings, in contrast, appear to lack such master memory maps, and instead retrieve information in far more haphazard fashion, by using clues (or cues) to what they're looking for, rather than knowing in advance where in the brain a given memory lies.” > **— Gary Marcus**, *2011, Edge Annual Question, “WHAT SCIENTIFIC CONCEPT WOULD IMPROVE EVERYBODY'S COGNITIVE TOOLKIT?”* [[reminders/Memory/Human Memory Lacks the Maps Computer Memory Uses by Gary Marcus|Human Memory Lacks the Maps Computer Memory Uses by Gary Marcus]]