Table 3 Summary of distinct characteristics in tabular-based transformer model.
From: Fault-class coverage–aligned combined training for AFDD of AHUs across multiple buildings
Aspect | TabTransformer | TabNet |
|---|---|---|
Attention mechanism | Full self-attention | Sparse feature attention |
Sequence handling efficiency | Moderate | High |
Feature selection capability | No | Yes |
Computational demand | Moderate | Lower |
Time series handling capability | Moderate | Strong |
