Publications by authors named "Panagiotis G Papaioannou"
Article Synopsis
- This study presents a three-tier numerical framework that uses nonlinear manifold learning to improve the forecasting of high-dimensional time series by addressing the challenges of high dimensionality during model training.
- The process includes three steps: embedding time series into a lower-dimensional space, constructing surrogate models for forecasting within that space, and lifting the forecasts back to the original high-dimensional space.
- The approach is tested on various problems, including synthetic time series related to EEG signals, solutions of linear and nonlinear PDEs, and a real-world dataset of foreign exchange rates from 2001 to 2020.
