Dynamical Reliability, Memory Retention and Information- Processing Capacity of a Six-Dimensional Series Memcapacitor System

Abstract This research examines whether the six-dimensional series-connected memcapacitor system can serve as a numerically reliable, input-driven computational substrate, and how its computational performance relates to the retention of its dynamical regime. A scalar input was added to the first state equation and the system was evaluated at six values of one coefficient, four coupling strengths, ten parameter, initial-state and solver scenarios per regime and five independent input realisations, giving 384 simulations. Lyapunov spectra were computed autonomously and along the driven trajectory, and memory, nonlinear and prediction performance were measured with ridge-regression readouts, a surrogate-based capacity threshold and reference predictors. At the published coefficients the leading exponent is λ₁ = 0.080 ± 0.011 (two standard errors) and the second exponent is indistinguishable from zero, so the regime is chaotic with a single positive exponent. This classification was retained in 9 of 10 and 8 of 10 perturbation scenarios in the two chaotic regimes and was lost only under ±5% perturbations, while the driven chaotic system escaped to infinity in four of five input realisations at the strongest coupling. Chaotic trajectories separated after 134 to 357 time units, whereas the Lyapunov estimates were insensitive to step size. Linear memory capacity reached 2.6 in non-chaotic regimes and was zero in the chaotic regimes and in a marginal non-chaotic regime. Nonlinear capacity was not found in the dynamics, and three of the six coordinates carried the useful information. NARMA-10 prediction (R² = 0.47 to 0.49) exceeded the mean but not a linear delay-line reference (0.66). Dynamical reliability and computational usefulness are therefore related but distinct. The results provide a quantitative framework for assessing memcapacitive systems as reservoirs, within the limits of a noiseless numerical model.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-16
DOI
https://doi.org/10.21203/rs.3.rs-11413535/v1
Primary Topic
Advanced Memory and Neural Computing
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Dynamical Reliability, Memory Retention and Information- Processing Capacity of a Six-Dimensional Series Memcapacitor System

I. O. AJAYI
2 links
Research Square
Advanced Memory and Neural Computing
preprint

Dynamical Reliability, Memory Retention and Information- Processing Capacity of a Six-Dimensional Series Memcapacitor System

I. O. AJAYI
preprint en

Abstract

Abstract This research examines whether the six-dimensional series-connected memcapacitor system can serve as a numerically reliable, input-driven computational substrate, and how its computational performance relates to the retention of its dynamical regime. A scalar input was added to the first state equation and the system was evaluated at six values of one coefficient, four coupling strengths, ten parameter, initial-state and solver scenarios per regime and five independent input realisations, giving 384 simulations. Lyapunov spectra were computed autonomously and along the driven trajectory, and memory, nonlinear and prediction performance were measured with ridge-regression readouts, a surrogate-based capacity threshold and reference predictors. At the published coefficients the leading exponent is λ₁ = 0.080 ± 0.011 (two standard errors) and the second exponent is indistinguishable from zero, so the regime is chaotic with a single positive exponent. This classification was retained in 9 of 10 and 8 of 10 perturbation scenarios in the two chaotic regimes and was lost only under ±5% perturbations, while the driven chaotic system escaped to infinity in four of five input realisations at the strongest coupling. Chaotic trajectories separated after 134 to 357 time units, whereas the Lyapunov estimates were insensitive to step size. Linear memory capacity reached 2.6 in non-chaotic regimes and was zero in the chaotic regimes and in a marginal non-chaotic regime. Nonlinear capacity was not found in the dynamics, and three of the six coordinates carried the useful information. NARMA-10 prediction (R² = 0.47 to 0.49) exceeded the mean but not a linear delay-line reference (0.66). Dynamical reliability and computational usefulness are therefore related but distinct. The results provide a quantitative framework for assessing memcapacitive systems as reservoirs, within the limits of a noiseless numerical model.

Research Square
Bamidele Olumilua University of Education, Science and Technology (NG)
Advanced Memory and Neural Computing
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon