Table 1 Framework’s basic hyperparameters
From: AutoTransOP: translating omics signatures without orthologue requirements using deep learning
Hyperparameter | L1000: 978 genes | L1000: 10,086 genes | Lung fibrosis | Serology |
|---|---|---|---|---|
Latent dimension | 292 | 1024 | 512 | 32 |
Hidden encoder layers dimensions | [640,384] | [4096,2048,1024,512] | [4096,2048,1024,512] | [64] |
Hidden decoder layers dimensions | [384,640] | [512,1024,2048, 4096] | [512,768,2048, 4096] | [64] |
Cell type classifier hidden layer dimensions | [256,128,64] | [512,256,128] | [256,128,64,32] | - |
Species classifier hidden layer dimensions | - | - | [256,128,64,32] | [32,16,8] |
Fibrosis classifier hidden layer dimensions | - | - | [256,128,64,32] | - |
Serology phenotype classifiers classifier hidden layer dimensions | - | - | - | [32,16,8] |
Adverse classifiers hidden layers dimensions | [256,128,64] | [512,256,128] | [512,256,128,64] | [32,16,8] |
Total batch size | 512 | 512 | 1024 | 50 |
Number of epochs | 1000 | 1000 | 200 | 2000 |
Learning rate | 0.001 | 0.001 | 0.001 | 0.001 |
