|
Notation | Explanation |
|
| A post from an online social media platform |
| -th character in a post |
| Prediction function |
| -dimensional matrix-vector |
| Number of classes contained in an episode |
| Number of instances in each class that are labeled |
| Number of instances in each class that are unlabelled |
| Vectorized representation of the -th character in a post |
| Query matrix-vector |
| Key matrix-vector |
| Value matrix-vector |
| Dimension of a query and key vector |
| Weight matrices corresponding to the -th header |
| Weight |
| Bias value |
| Storage cell state of BiLSTM |
| Temporary cell state of BiLSTM |
| Hidden state of BiLSTM |
| Forgetting gate of BiLSTM |
| Memory gate of BiLSTM |
| Output gate of BiLSTM |
| Logistic regression function |
| hidden states of the whole BiLSTM |
HD | Historical rumor dataset |
ED | Epidemic rumor dataset |
| Hidden state |
| Support set vector |
| Query set vector |
| Number of neurons in the output capsule |
| Number of neurons (vector length) in each capsule |
| Prediction vector in the Capsnet |
| Transformation matrix in the Capsnet |
| Connection weight in the Capsnet |
| Coupling coefficient in the Capsnet |
| Output of the -th upper capsule in the Capsnet |
| General representation of the -th historical rumor class |
| Epidemic rumor query sample |
| A sample of the same category as |
| A sample of a different category from |
|