ETA Prediction with Graph Neural Networks in Google Maps - Paper Explained

Dive into a comprehensive video explanation of the paper "ETA Prediction with Graph Neural Networks in Google Maps." Explore how Google Maps utilizes Graph Neural Networks (GNNs) for real-world applications, including graph formation, feature extraction, and the DeepMind GN model. Learn about different prediction horizons, loss functions, variance reduction techniques, and ETA baselines. Understand the inference process, offline results, and various experiments conducted. Gain insights into the engineering aspects of implementing GNNs in production for accurate travel time estimation.

Intro - GNNs in production
How graphs are formed
Graph features
GNN explained DeepMind GN
Different horizons
Loss functions
Reducing the variance
ETA baselines explained
How does the inference work
Offline results
Ablations and experiments
Outro, engineering