Skip to main content

Table 1 The number of model parameters and total training time for foreground models in 30 iterations

From: Sequence-specific bias correction for RNA-seq data using recurrent neural networks

  GRU-10 GRU-20 LSTM-10 LSTM-20
Number of parameters 524 1644 684 2164
Training time 375.1 s 529.3 s 361.7 s 526.3 s
Perplexity of testing set 3.833 3.823 3.833 3.822
  1. The training time is calculated on a 2.8 GHz CPU Macbook laptop where only the CPU is used