1Background
Transformer relies entirely on attention and drops recurrence [1], built on scaled dot-product attention:
Attention(Q, K, V) = softmax(Q KT√dk) V
Traffic on neighboring road segments is correlatedLin, which is exactly what attention is good at.
2Method
@Nuo write the method section and put the citations at the end
We apply attention across space and time to capture how road segments relate and how traffic cycles, then fuse the two for forecasting.Nuo
3Schedule
TaskOwnerStatus
Literature reviewNuoDone
Data cleanupLinIn progress
Baseline runChenNot started