Table 2 Details of network architecture.
From: Newly reconstructed Arctic surface air temperatures for 1979–2021 with deep learning method
Module Name | Filter Size | Channels | Stride | Padding | Nonlinearity |
|---|---|---|---|---|---|
PConv1 | 7 × 7 | 18 | 2 | 3 | ReLU |
PConv2 | 5 × 5 | 36 | 2 | 2 | ReLU |
PConv3 | 5 × 5 | 72 | 2 | 2 | ReLU |
UpSample1 | — | 72 | — | — | — |
Concat1 | — | 72 + 36 | — | — | — |
PConv4 | 3 × 3 | 36 | 1 | 1 | LeakyReLU |
UpSample2 | — | 36 | — | — | — |
Concat2 | — | 36 + 18 | — | — | — |
PConv5 | 3 × 3 | 18 | 1 | 1 | LeakyReLU |
UpSample3 | — | 18 | — | — | — |
Concat3 | — | 18 + 3 | — | — | — |
PConv6 | 3 × 3 | 3 | 1 | 1 | LeakyReLU |
