Online Class-Incremental Learning on Split-MNIST
94Final Mean AccuracySQDA
Evaluation Results
| Method | Links | |
|---|---|---|
| SQDAType=Non-tree baseline2026.05 | 94 | |
| SQDAProtocol=Online CIL (single-pass)2026.05 | 94 | |
| PECType=Exemplar-free baseline2026.05 | 89.8 | |
| PECProtocol=Online CIL (single-pass)2026.05 | 89.8 | |
| VFDTProtocol=No-reset CIL-adapted2026.05 | 87.4 | |
| VFDTProtocol=Online CIL, Reset Variant=No-reset, Data Retention=None, Number of passes=Single-pass2026.05 | 87.4 | |
| MIST-GProtocol=Online CIL2026.05 | 86.3 | |
| MIST-GProtocol=Online CIL (single-pass)2026.05 | 86.3 | |
| MIST-GProtocol=Online CIL, Reset Variant=No-reset, Data Retention=None, Number of passes=Single-pass2026.05 | 86.3 | |
| SLDAType=Non-tree baseline2026.05 | 85.4 | |
| SLDAProtocol=Online CIL (single-pass)2026.05 | 85.4 | |
| RutkowskiProtocol=No-reset CIL-adapted2026.05 | 85.1 | |
| RutkowskiProtocol=Online CIL, Reset Variant=No-reset, Data Retention=None, Number of passes=Single-pass2026.05 | 85.1 | |
| RLSType=Non-tree baseline2026.05 | 81.8 | |
| RLSProtocol=Online CIL (single-pass)2026.05 | 81.8 | |
| NCMType=Non-tree baseline2026.05 | 79.4 | |
| NCMProtocol=Online CIL (single-pass)2026.05 | 79.4 | |
| ARFTrees=10, Drift detection=Disabled2026.05 | 78.2 | |
| ARF-NRProtocol=Online CIL, Reset Variant=No-reset, Data Retention=None, Number of passes=Single-pass, Ensemble Size=10 trees, Drift Detection=Disabled2026.05 | 78.2 | |
| MIST-KProtocol=Online CIL2026.05 | 77.2 | |
| MIST-KProtocol=Online CIL (single-pass)2026.05 | 77.2 | |
| MIST-KProtocol=Online CIL, Reset Variant=No-reset, Data Retention=None, Number of passes=Single-pass2026.05 | 77.2 | |
| HATVariant=No-reset (NR)2026.05 | 32.4 | |
| HAT (NR)Protocol=Online CIL, Reset Variant=No-reset, Data Retention=None, Number of passes=Single-pass2026.05 | 32.4 | |
| EFDTVariant=No-reset (NR)2026.05 | 18.2 | |
| EFDT (NR)Protocol=Online CIL, Reset Variant=No-reset, Data Retention=None, Number of passes=Single-pass2026.05 | 18.2 |