Accelerating Simulation of Stiff Nonlinear Systems using Continuous-Time Echo State Networks

The paper defines CTESNs with r′ = tanh(Ar + Win x(p*, t)) and x(t) = Wout r(t), trains Wout by least-squares on the same reservoir time series for all training parameters, interpolates Wout(p) via RBFs across parameter space, and emphasizes that the reservoir has no parameter dependence, so its trajectory can be reused at prediction time; it also reports near-constant-time surrogate evaluation . The candidate solution reproduces these claims and fills in standard linear-algebra details (normal equations, Moore–Penrose solution, conditions for uniqueness) and an explicit RBF construction and complexity estimate. The paper is correct on these points but does not formalize the LS solution or provide tight complexity bounds (which it explicitly notes) , whereas the model supplies a precise derivation and cost accounting. Hence both are correct, with the model offering a more explicit proof/derivation.