Continuous Time Event Sequence

Continuous-time event sequences (CTES) model data where events occur at irregular intervals over continuous time, a common pattern in diverse fields like finance, healthcare, and social media. Current research focuses on developing efficient and robust models, such as neural ordinary differential equations (ODEs), transformers adapted for continuous time (e.g., Rough Transformers), and neural temporal point processes (TPPs), to capture complex temporal dependencies and handle issues like missing data and streaming data. These advancements are improving the accuracy and scalability of applications ranging from event prediction and sequence retrieval to personalized recommendations and causal structure learning from event data.