Latent SDEs for Anomaly Detection in Sparse Multivariate Time Series

We propose a generative method using Latent SDEs to detect anomalies in sparse and irregular multivariate time series. The approach projects observed data onto continuous-time stochastic systems, handling missing values and irregular sampling while capturing cyclic patterns. Experiments on six benchmark datasets show our method achieves top performance, outperforming state-of-the-art baselines, especially under severe data sparsity.