Neural CDEs

Neural Controlled Differential Equations (Neural CDEs) are a class of neural networks designed to model dynamical systems from data, particularly time series with irregular sampling. Current research focuses on improving the robustness and efficiency of Neural CDEs, exploring variations like Log-NCDEs and invertible Neural CDEs, and applying them to diverse problems including drug-drug interaction prediction, image reconstruction, and control systems. This approach offers advantages over traditional methods by handling complex temporal dependencies and leveraging the continuous nature of underlying physical processes, leading to improved accuracy and efficiency in various scientific and engineering applications.