Table 4.
Brief comparison of methods for addressing non-ideal environment
| Methods | Explainability | Resource consumption | Challenges |
|---|---|---|---|
| Preprocessing | Strong | Low | Possibility of losing useful feature information. |
| Deep learning | Weaka | High (computing resources) | High requirements for neural network design. |
| Data augmentation | Median | High (storage resources) | Some methods rely on known ideal signal. |
Note: (a)Due to the lack of interpretability [82] of neural networks.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.