Figures & data
Table 1. Common activation functions.
Figure 2. Different ANN model architectures: (a) simple feed-forward neural network, (b) a recurrent (Elman) neural network, and (c) a deep feed-forward neural network with multiple hidden layers.
![Figure 2. Different ANN model architectures: (a) simple feed-forward neural network, (b) a recurrent (Elman) neural network, and (c) a deep feed-forward neural network with multiple hidden layers.](/cms/asset/61373050-5d46-4645-af04-269bd7792480/uawm_a_1459956_f0002_b.gif)
Table 2. Enhanced first-order optimizers used for DL.
Figure 5. LSTM architecture showing unit time delays (−1), gates, and recurrentactivation functions (σ).
![Figure 5. LSTM architecture showing unit time delays (−1), gates, and recurrentactivation functions (σ).](/cms/asset/cd8956d1-753e-4045-ab7c-247c9d7d3d85/uawm_a_1459956_f0005_b.gif)
Table 3. Chemical and meteorological parameters captured at the Air Monitoring Station.
Figure 11. Feature importance from decision tree prediction of 8-hr O3 exceedances from 1 hr to 12 hr.
![Figure 11. Feature importance from decision tree prediction of 8-hr O3 exceedances from 1 hr to 12 hr.](/cms/asset/e4a931a4-06ea-4f98-a0b9-cb195bc0ec6f/uawm_a_1459956_f0011_b.gif)
Figure 13. Loss function errors for training and test data sets for different horizons at (a) 24 hr, (b) 36 hr, and (c) 48 hr.
![Figure 13. Loss function errors for training and test data sets for different horizons at (a) 24 hr, (b) 36 hr, and (c) 48 hr.](/cms/asset/5754c768-1584-4b7d-9e7f-a4edc3643e21/uawm_a_1459956_f0013_b.gif)
Table 4. Default values for parameter sensitivity analysis.
Figure 18. Impact of (a) batch samples, (b) look-back nodes, and (c) dropout factor parameters on training errors in the model.
![Figure 18. Impact of (a) batch samples, (b) look-back nodes, and (c) dropout factor parameters on training errors in the model.](/cms/asset/dccafc9b-6aeb-4a24-918a-31c029150aec/uawm_a_1459956_f0018_b.gif)
Table 5. Comparison of RNN test data results to previously published results.
Table 6. Comparison of different forecasting errors over a 24-hr period.