Delay Distribution

Delay distribution research focuses on understanding and mitigating the impact of delayed feedback in various sequential decision-making problems, such as online advertising and reinforcement learning. Current research explores algorithms, including Thompson Sampling and optimistic approaches, designed to handle stochastic delays with varying distributions, even those with unbounded expectations, and utilizes models like mixture density networks and graph neural networks for accurate delay prediction in specific applications. This work is crucial for improving the efficiency and reliability of systems operating under real-world conditions where immediate feedback is often unavailable, impacting fields ranging from wireless networking to datacenter management.