Minimal descriptions of cyclic memories
read the original abstract
Many materials that are out of equilibrium can "learn" one or more inputs that are repeatedly applied. Yet, a common framework for understanding such memories is lacking. Here we construct minimal representations of cyclic memory behaviors as directed graphs, and we construct simple physically-motivated models that produce the same graph structures. We show how a model of worn grass between park benches can produce multiple transient memories---a behavior previously observed in dilute suspensions of particles and charge-density-wave conductors---and the Mullins effect. Isolating these behaviors in our simple model allows us to assess the necessary ingredients for these kinds of memory, and to quantify memory capacity. We contrast these behaviors with a simple Preisach model that produces return-point memory. Our analysis provides a unified method for comparing and diagnosing cyclic memory behaviors across different materials.
This paper has not been read by Pith yet.
Forward citations
Cited by 1 Pith paper
-
Learning by training: emergent return-point memory from cyclically tuning disordered sphere packings
Cyclic inverse design on athermal disordered sphere packings produces an emergent marginally absorbing manifold that encodes return-point memory of the training range through gradient discontinuities.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.