Skip to main content

Table 4 Training details on NLP models. “Max. It.” means maximum number of iterations allowed

From: Towards a robust out-of-the-box neural network model for genomic data

Model Data Max. It. Patience Early stop Optimizer
LSTM-AE (32) Splice 2000 100 1474 Adam (LR=0.001)
LSTM-AE+NN (32) Splice 1000 100 123 SGD (LR=0.01)
LSTM-AE (256) Splice 2000 100 424 Adam (LR=0.001)
LSTM-AE+NN (256) Splice 1000 100 114 SGD (LR=0.01)
LSTM-AE (1024) Splice 2000 100 1005 Adam (LR=0.001)
LSTM-AE+NN (1024) Splice 1000 100 200 SGD (LR=0.01)
LSTM-layer Splice 4000 100 289 Adam (LR=0.001)
LSTM-layer Splice 4000 100 2872 SGD (LR=0.01)
doc2vec+NN (50) Splice 1000 50 119 SGD (LR=0.01)
doc2vec+NN (100) Splice 1000 50 143 SGD (LR=0.01)
doc2vec+NN (150) Splice 1000 50 166 SGD (LR=0.01)
doc2vec+NN (200) Splice 1000 50 264 SGD (LR=0.01)
LSTM-AE (32) Histone 4000 100 549 Adam (LR=0.001)
LSTM-AE+NN (32) Histone 1000 100 212 SGD (LR=0.001)
LSTM-AE (256) Histone 4000 200 422 Adam (LR=0.001)
LSTM-AE+NN (256) Histone 1500 200 805 SGD (LR=0.001)
LSTM-AE (1024) Histone 4000 200 646 Adam (LR=0.001)
LSTM-AE+NN (1024) Histone 1500 200 999 SGD (LR=0.001)
LSTM-layer Histone 3000 100 154 Adam (LR=0.001)
LSTM-layer Histone 4000 400 526 SGD (LR=0.01)
doc2vec+NN (50) Histone 1000 30 114 SGD (LR=0.01)
doc2vec+NN (100) Histone 1000 30 177 SGD (LR=0.01)
doc2vec+NN (150) Histone 1000 30 61 SGD (LR=0.01)
doc2vec+NN (200) Histone 1000 30 118 SGD (LR=0.01)
LSTM-AE (32) Motif 200 10 78 Adam (LR=0.001)
LSTM-AE+NN (32) Motif 500 10 185 SGD (LR=0.01)
LSTM-AE (256) Motif 200 10 195 Adam (LR=0.001)
LSTM-AE+NN (256) Motif 500 10 146 SGD (LR=0.01)
LSTM-AE (1024) Motif 200 10 200 Adam (LR=0.001)
LSTM-AE+NN (1024) Motif 500 10 62 SGD (LR=0.01)
LSTM-layer Motif 200 5 51 Adam (LR=0.001)
LSTM-layer Motif 200 5 200 SGD (LR=0.01)
doc2vec+NN (50) Motif 400 10 36 SGD (LR=0.01)
doc2vec+NN (100) Motif 400 10 39 SGD (LR=0.01)
doc2vec+NN (150) Motif 400 10 58 SGD (LR=0.01)
doc2vec+NN (200) Motif 400 10 11 SGD (LR=0.01)