Parsimonious description for predicting high-dimensional dynamics.

Area of Science:

• Dynamical Systems
• Data Science
• Time Series Analysis

Background:

• Observing complex systems often yields limited measurements, necessitating methods to reconstruct the full system state.
• Delay coordinates, formed from successive measurements, are a standard technique for state reconstruction in dynamical systems.
• Classical delay coordinate methods face computational challenges in high-dimensional systems.

Purpose of the Study:

• To develop a computationally efficient and accurate method for reconstructing high-dimensional dynamical systems from limited measurements.
• To address the limitations of classical delay coordinate methods in terms of computational cost and bias.
• To improve the prediction of future system states using a parsimonious description of delay coordinates.

Main Methods:

• Proposed a parsimonious description of delay coordinates using exponentially decaying weights.
• Evaluated distances between delay vectors with these decaying weights.
• Applied the method to toy models of atmospheric dynamics and real-world renewable energy datasets.

Main Results:

• The proposed method allows for faster prediction of future measurement values by reusing calculated distances.
• The approach provides more accurate predictions by reducing bias towards stable system directions.
• Demonstrated effectiveness on both simulated and real-world complex datasets.

Conclusions:

• The novel delay coordinate description offers a computationally efficient and accurate alternative for state reconstruction in high-dimensional dynamical systems.
• This method enhances predictive capabilities, particularly in fields like atmospheric science and renewable energy.
• The parsimonious approach mitigates biases inherent in traditional delay coordinate techniques.