Table 1 Comparison of advantages and disadvantages of individual technologies.
From: Unbalanced power anomaly detection model based on improved transformer and countermeasure encoder
Type of technology | Advantages | Limitations |
|---|---|---|
Time-Transformer (2023) | Parallel training is supported, and the inference speed is 3.2 times faster than that of LSTM on ECG5000 data set. | Memory footprint grows with the square of the sequence length |
GAN-SMOTE (2022) | The Diversity Index of the generated sample is 0.81, which is 29% higher than that of the traditional SMOTE. | Training convergence requires 2000 + iterations |
GraphSleepNet (2024)20 | Through node embedding to capture device correlation, the F1-score of multi-meter joint detection is increased to 0.887. | Rely on a predefined topology |
