Metropolis-Scale Resilient and Trustworthy Traffic Flow Inference Using Multi-Source Data

Name: Metropolis-Scale Resilient and Trustworthy Traffic Flow Inference Using Multi-Source Data
Uploaded: 2026-10-23T09:59:41.120810+00:00
Duration: 914 s
Description: The paper introduces Task-Aware Attentive Neural Process (TA-ANP) to accurately infer network traffic states using sparse data from floating cars and fixed detectors. It solves multi-task interference by using task-aware attention and models uncertainty via Monte Carlo Dropout, leading to state-of-the-art performance and resilient predictions.

The paper introduces Task-Aware Attentive Neural Process (TA-ANP) to accurately infer network traffic states using sparse data from floating cars and fixed detectors. It solves multi-task interference by using task-aware attention and models uncertainty via Monte Carlo Dropout, leading to state-of-the-art performance and resilient predictions.

Open the full episode