Related Experiment Video
Updated: Jun 24, 2025

Visualizing Hyporheic Flow Through Bedforms Using Dye Experiments and Simulation
Published on: November 18, 2015
Hybridizing traditional and next-generation reservoir computing to accurately and efficiently forecast dynamical
R Chepuri1, D Amzalag2, T M Antonsen1,3,4
1Department of Physics, University of Maryland, College Park, Maryland 20742, USA.
A new hybrid approach combines reservoir computers (RCs) and next-generation reservoir computers (NGRCs) for improved time series prediction. This method enhances accuracy and efficiency, especially for chaotic dynamical systems with limited data or resources.
Area of Science:
- Computational Science
- Machine Learning
- Dynamical Systems
Background:
- Reservoir computers (RCs) excel at time series prediction.
- Next-generation reservoir computers (NGRCs) offer computational and data efficiency.
- NGRCs face challenges with sampling sensitivity and data nonlinearities.
Purpose of the Study:
- To introduce a hybrid RC-NGRC approach for time series forecasting.
- To evaluate its performance on chaotic dynamical systems.
- To address limitations of standalone RC and NGRC methods.
Main Methods:
- Developed a hybrid model integrating RC and NGRC architectures.
- Tested the hybrid approach on multiple model chaotic systems.
- Compared performance against traditional RCs and NGRCs under resource constraints.
Main Results:
- The hybrid RC-NGRC approach accurately predicts short-term dynamics and captures long-term statistics of chaotic systems.
- It outperforms standalone RCs and NGRCs when computational resources are limited or data is sparse.
- Achieved prediction performance comparable to large RCs with a smaller reservoir size.
Conclusions:
- The hybrid RC-NGRC method offers significant computational efficiency gains over traditional RCs.
- It effectively mitigates NGRC limitations, particularly in challenging forecasting scenarios.
- This approach is highly beneficial when computational efficiency is critical and NGRCs alone are insufficient.
Related Concept Videos
Multi-input and Multi-variable systems
In the absence...
Design Example: Analyzing Capacity Contours for Flood Risk Assessment
Hybrid Zones
Cyclic Processes And Isolated Systems
In the case of a non-isolated system, the change in the internal energy is zero only if the process is cyclic. A thermodynamic process is considered cyclic if the system undergoes a series of changes and returns to its initial state.
Consider a cyclic process that returns to its initial state, undergoing a four-step process. The heat transfer along each...
Design Example: Creating a Hydraulic Model of a Dam Spillway
Distributed Loads: Problem Solving

