Air Quality Prediction Model Based on Spatiotemporal Data Analysis and Metalearning

Algorithm 1

MetaGAT-LSTM.

Input:: The set of the training datasets from source cities;

   : Datasets from target city;

   : Distribution over ;

: Learning rates

Output: : The GAT-LSTM model for the target city

1. Randomly initialize

2. While not done do:

3.  Sample datasets from with replacement w.r.t.

4.  Get next training batches from , respectively

5.  Get next training batch from

6.  For in do:

7.   Calculate with respect to

8.   Calculate first-step adapted parameter

9.   Calculate with respect to

10.  End for

11.  Calculate second-step adapted parameter

12. End while