EOMR: A New Method for Joint Feature Subset and Subspace Learning in Regression

A new extension to the Entropy-Optimal Manifold Clustering (EOMC) method, called Entropy-Optimal Manifold Regression (EOMR), has been introduced in a recent arXiv paper (arXiv:2607.28080). The method allows for joint simultaneous identification of relevant feature subsets and subspaces in nonstationary and nonlinear regression problems. EOMR is designed to achieve robust learning with linearly-scaling iteration and memory complexities. The authors compared EOMR against a comprehensive set of state-of-the-art AI and ML tools on challenging problems from chaotic and fluid dynamics, including predicting the Lorenz-96 system in strongly and very-strongly chaotic regimes (with forcing parameters F=8 and F=12) and data from the Hasegawa-Wakatani model on the edge of tokamak plasma. The benchmarks demonstrate the effectiveness of EOMR in these contexts.