Hierarchical ODE: Learning Continuous-Time Physical Prototypes for Early Link Failure Detection

The article presents a hierarchical ordinary differential equation (ODE) clustering network designed for time series prototype learning, specifically aimed at early link failure detection. This approach utilizes neural ODEs to model latent state evolution as continuous integral curves, allowing for the effective separation of smooth feature trends from stochastic noise, while an adaptive mechanism determines the number of prototypes dynamically. This methodology is significant for practitioners as it enhances robustness in failure detection from irregularly sampled time series, addressing limitations of traditional discrete architectures.