Novelty Detection in Time Series via Weak Innovations Representation: A Deep Learning Approach
This addresses novelty detection for time series analysis, but it appears incremental as it builds on existing innovations representation concepts with a deep learning twist.
The paper tackles novelty detection in time series with unknown probability structures by proposing a deep learning method to extract an innovations sequence, achieving minimax optimality under a Bayes risk measure and demonstrating robustness in experiments.
We consider novelty detection in time series with unknown and nonparametric probability structures. A deep learning approach is proposed to causally extract an innovations sequence consisting of novelty samples statistically independent of all past samples of the time series. A novelty detection algorithm is developed for the online detection of novel changes in the probability structure in the innovations sequence. A minimax optimality under a Bayes risk measure is established for the proposed novelty detection method, and its robustness and efficacy are demonstrated in experiments using real and synthetic datasets.