Conceptualizations of 'memory' - more elusive than you might think
Year: 2024
Type: article
Abstract: Work presentation at the CogniGron center for neuromorphic computing, September 16th, 2024 Abstract The term 'memory' is easily used in many of the discussions in our multidisciplinary center (physics, AI, mathematics, computer science, cognitive science). The advent of computers has led to a dominant interpretation of what 'memory' is about. Still, in a multidisciplinary setting, many ambiguities are present. In this presentation I will present a number of views and implementations of 'memory' and propose to introduce the notion of functional memory, i.e., information-processing systems that give a consistent response when probed, however, without requiring that the internal mechanics are perfect and consistent at the level of the individual device or connection. Notably there is a role for energy-based systems, where energy minima correspond to to stored vectorial states. The Hopfield/Ising model is but one of the possible realizations and it is hoped that more powerful variants will be developed in the near future. An interesting aspect of energy-based models allow for the characterization of decision processes in memory by looking at the detailed shape of the landscape. In the presentation, examples are given, e.g., of a stubborn system condition as compared to the 'push over', where just a little bit of energy is needed to jump to another state. MP4 of powerpoint presentation with speaker on cam.
Author: Schomaker, Lambert
Institution: University of Groningen
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6967966460
FWCI:
Cites:
Cited by:
Related to: