Table 2 Input/output dimensions of each layer of proposed model
From: Sentiment analysis from textual data using multiple channels deep learning models
Layers | Input shape | Output shape |
|---|---|---|
Input shape (with sequence length = 300) | (None, 300) | (None, 300) |
Embedding | (None, 300) | (None, 300, 300) |
Convolutional layer (for each layer out of 5) | (None, 300, 300) | (None, 300,128) |
BiLSTM layer I | (None, 300, 128) | (None, 300, 64) |
BiLSTM layer II | (None, 300, 64) | (None, 300, 64) |
Batch normalization | (None, 300, 64) | (None, 300, 64) |
Attention layer | (None, 300, 64) | (None, 64) |
Dense layer | (None, 64) | (None, 128) |
Concatenation of five dense layers for each channel | (None, 128) * 5 | (None, 640) |
Dense layer | (None, 640) | (None, 16) |
Dropout layer | (None, 16) | (None, 16) |
Output layer | (None, 16) | (None, 1) |