Skip to main content
. 2022 Jan 22;41:219–231. doi: 10.1016/j.jare.2022.01.009

Fig. 1.

Fig. 1

Optimization of hyperparameters and architectures for CNN-RNN (BiLSTM) model with dictionary encoding method. (A) Performance comparison of different filter and kernel size combinations; (B) Performance comparison of different pooling sizes; (C) Performance comparison of different numbers of convolutional and pooling layers; (D) Performance comparison of different RNN architecture types; (E) Performance comparison of different numbers of hidden cells in the bidirectional LSTM layer; (F) Schematic illustration of the optimized model architecture.