Table 26 Computational cost and inference time comparison across models.
From: PatternFusion: a hybrid model for pattern recognition in time-series data using ensemble learning
Model configuration | Parameter count (M) | Inference time (ms) | FLOPs (GFLOPs) |
|---|---|---|---|
VGGFace2 + IrisCode | 42.5 | 56 | 120.8 |
CNN-LSTM Fusion | 35.8 | 49 | 105.3 |
Multi-stream transformers | 60.3 | 72 | 140.5 |
Informer | 48.7 | 65 | 130.2 |
Autoformer | 50.1 | 68 | 135.6 |
PatternFusion (proposed) | 28.4 | 42 | 98.7 |
