A Mixture-of-Experts Framework for Multi-Stream Quickest Change Detection

Name: A Mixture-of-Experts Framework for Multi-Stream Quickest Change Detection
Uploaded: 2026-10-20T11:00:54.487557+00:00
Duration: 445 s
Description: This work proposes a Mixture-of-Experts (MoE) framework for multi-stream Quickest Change Detection (QCD). It combines statistics from multiple data streams using weighted sums, outperforming traditional methods like the rank-weighted Shiryaev–Roberts scheme. The findings show that the best way to weight these experts depends on how many streams are affected by a change.

This work proposes a Mixture-of-Experts (MoE) framework for multi-stream Quickest Change Detection (QCD). It combines statistics from multiple data streams using weighted sums, outperforming traditional methods like the rank-weighted Shiryaev–Roberts scheme. The findings show that the best way to weight these experts depends on how many streams are affected by a change.

Open the full episode