CNFRD: A Few-Shot Rumor Detection Framework via Capsule Network for COVID-19

Table 4

The specific experimental parameters setting.

Parameter	Parameter value	Parameter explanation
Lstm_dim	128	Hidden layer dimension of LSTM
Iterations	3	Iterations are used by the routing algorithm
Seed	42	Random seed
Margin_loss_scaling		Scale margin loss
Dropout	0.1	Drop rate
Adam_epsilon	1 e − 8	Epsilon for adam optimizer
Train_batch_size	1000	Total batch size for training
Eval_batch_size	84	Total batch size for eval
Test_batch_size	84	Total batch size for test
Num_train_epochs	50	Total number of training epochs to perform
Learning_rate	3 e − 6	The initial learning rate for adam
Max_seq_length	164	The maximum length of the rumor text sequence