{"ID":2890947,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2507.18467","arxiv_id":"2507.18467","title":"Contraction, Criticality, and Capacity: A Dynamical-Systems Perspective on Echo-State Networks","abstract":"Echo-State Networks (ESNs) distil a key neurobiological insight: richly recurrent but fixed circuitry combined with adaptive linear read-outs can transform temporal streams with remarkable efficiency. Yet fundamental questions about stability, memory and expressive power remain fragmented across disciplines. We present a unified, dynamical-systems treatment that weaves together functional analysis, random attractor theory and recent neuroscientific findings. First, on compact multivariate input alphabets we prove that the Echo-State Property (wash-out of initial conditions) together with global Lipschitz dynamics necessarily yields the Fading-Memory Property (geometric forgetting of remote inputs). Tight algebraic tests translate activation-specific Lipschitz constants into certified spectral-norm bounds, covering both saturating and rectifying nonlinearities. Second, employing a Stone-Weierstrass strategy we give a streamlined proof that ESNs with polynomial reservoirs and linear read-outs are dense in the Banach space of causal, time-invariant fading-memory filters, extending universality to stochastic inputs. Third, we quantify computational resources via memory-capacity spectrum, show how topology and leak rate redistribute delay-specific capacities, and link these trade-offs to Lyapunov spectra at the \\textit{edge of chaos}. Finally, casting ESNs as skew-product random dynamical systems, we establish existence of singleton pullback attractors and derive conditional Lyapunov bounds, providing a rigorous analogue to cortical criticality. The analysis yields concrete design rules-spectral radius, input gain, activation choice-grounded simultaneously in mathematics and neuroscience, and clarifies why modest-sized reservoirs often rival fully trained recurrent networks in practice.","short_abstract":"Echo-State Networks (ESNs) distil a key neurobiological insight: richly recurrent but fixed circuitry combined with adaptive linear read-outs can transform temporal streams with remarkable efficiency. Yet fundamental questions about stability, memory and expressive power remain fragmented across disciplines. We present...","url_abs":"https://arxiv.org/abs/2507.18467","url_pdf":"https://arxiv.org/pdf/2507.18467v1","authors":"[\"Pradeep Singh\",\"Lavanya Sankaranarayanan\",\"Balasubramanian Raman\"]","published":"2025-07-24T14:41:18Z","proceeding":"cs.NE","tasks":"[\"cs.NE\",\"nlin.CD\"]","methods":"[]","has_code":false}