Table 7 Comparison of computation time across approachs with different layers and architectures.

From: A hybrid deep learning air pollution prediction approach based on neighborhood selection and spatio-temporal attention

Approach

LSTM

CNN-LSTM

STA-ResCNN

STARes-LSTM

KSC-ConvLSTM

Training step time (ms/step)

9.7

278.8

339.7

490.8

254.3

Back to article page