Table 2 Hyperparameter values and setting for MSW-ViT.
From: AI meets endodontics a deep learning approach to precision diagnosis
Parameters | Value / Settings |
|---|---|
Learning rate | [10− 2 to10− 4] |
Feature vector size | Problem specific |
Self-attention window size | 5 × 5 |
No of heads | 12 |
Depth rate | 0.3 |
Dropout rate | 0.4 |
Training: Testing | 80: 20 |
Embedding dimension | 256 |
Attention mechanism | Shifted window |
Batch size | 256 |
Scheduler | StepLR |
No of Epochs | 60 |
Option | Default |
