m-WCN Framework for Joint Temporal and Spectral Time Series Analysis
September 25, 2026
The m-WCN framework neuralizes multi-wavelet decomposition by using trainable convolutional operators and orthogonality constraints to replace classical GHM transforms. This enables end-to-end joint extraction of temporal patterns and frequency components for improved forecasting and classification.
HOW THIS AFFECTS YOU
●
researcherYou can use this method to capture multiscale characteristics in time series without decoupling frequency and time-domain modeling.