Table 4 Parameter configuration for each algorithm.

From: Hybrid-driven modeling using a BiLSTM–AdaBoost algorithm for diameter prediction in the constant diameter stage of Czochralski silicon single crystals

Algorithm

Parameter settings

CNN

Kernel size = 3, Number of kernels = 128, Pooling window = 2, Dropout = 0.2, Learning rate = 0.001

LSTM

Hidden units = 128, Number of layers = 1, Dropout = 0.2, Learning rate = 0.001

GRU

Hidden units = 128, Number of layers = 1, Dropout = 0.2, Learning rate = 0.001

CNN-BiLSTM

Kernel size = 3, Number of kernels = 128, Pooling window = 2, BiLSTM hidden units = 128, Layers = 1, Dropout = 0.2, Learning rate = 0.001

CNN-LSTM

Kernel size = 3, Number of kernels = 128, Pooling window = 2, LSTM hidden units = 128, Layers = 1, Dropout = 0.2, Learning rate = 0.001

BiLSTM-AdaBoost

BiLSTM hidden units = 128, Layers = 1, Dropout = 0.2, AdaBoost weak classifiers = 50, Learning rate = 0.001

Back to article page