A Sparse Gating Convolutional Recurrent Network for Traffic Flow Prediction

Table 5

The details of the module on the TaxiNYC dataset.

Hyperparameter	ConvGRU	SConvGRU	SConvGRU+	ConvLSTM	SConvLSTM	SConvLSTM+
Kernel size in CNN	(3, 5)	(3, 5)	(3, 5)	(3, 5)	(3, 5)	(3, 5)
Stride in CNN	1 (first layer), 1 (second layer)
Unit layers	2	2	2	2	2	2
Hidden channels of unit	32 (first layer), 64 (second layer)			32 (first layer), 32 (second layer)
Kernel size in DCNN	(3, 5)	(3, 5)	(3, 5)	(3, 5)	(3, 5)	(3, 5)
Stride in DCNN	1 (first layer), 1 (second layer)