The field of time series analysis is moving towards the development of innovative methods to handle irregularly sampled and incomplete data. Recent advances focus on integrating stochastic modeling with neural networks to capture delayed temporal dynamics and handle missing values. This direction enables the effective classification and detection of novel events in astronomical and industrial systems. Noteworthy papers include Modeling Irregular Astronomical Time Series with Neural Stochastic Delay Differential Equations, which introduces a principled framework for time series analysis under observational constraints. TANDEM and STDiff also demonstrate superior performance in handling missing data in time series classification, while ZTFed-MAS2S provides a secure and efficient solution for wind power data imputation using zero-trust federated learning.