Table 3 Details of autoencoder architectures. Here AE* indicates the number of layers in compression autoencoders, and AE*_L1 indicates the number of layers in L1-constrained autoencoders
From: Overlapping community finding with noisy pairwise constraints
Architecture | Nodes per layer | Small networks | Large networks | ||
|---|---|---|---|---|---|
Epochs | Learning rate | Epochs | Learning rate | ||
AE1 | dim:(7,3,7) | 100 | 0.01 | 30 | 0.001 |
AE1_L1 | dim:(7,7,7) | 100 | 0.01 | 30 | 0.001 |
AE2 | dim:(7,5,3,5,7) | 100 | 0.01 | 30 | 0.001 |
AE2_L1 | dim:(7,7,7,7,7) | 100 | 0.01 | 30 | 0.001 |
AE3 | dim:(7,6,5,3,5,6,7) | 100 | 0.01 | 30 | 0.001 |
AE3_L1 | dim:(7,7,7,7,7,7,7) | 100 | 0.01 | 30 | 0.001 |