Echo-State Networks Reproduce Rare Events in Chaotic Systems
In a study, researchers utilized Echo-State Networks to forecast time series and statistical characteristics of the Lotka-Volterra model under chaotic conditions. These networks effectively captured the chaotic attractor and generated histograms for dependent variables, encompassing tails and infrequent occurrences. Additionally, they were able to replicate rare events during non-equilibrium simulations of the Lotka-Volterra system. To assess tail behavior, the Generalized Extreme Value distribution was employed.
Key facts
- Echo-State Networks were applied to the competitive Lotka-Volterra model.
- The model operates in the chaotic regime.
- Networks learned the chaotic attractor.
- Histograms of dependent variables were reproduced, including tails and rare events.
- Rare events in non-equilibrium simulations were also reproduced.
- Generalized Extreme Value distribution quantified tail behavior.
- The research is from Nonlinear Sciences > Chaotic Dynamics.
- The paper is available on arXiv with ID 2505.16208.
Entities
Institutions
- arXiv