Parkinson's Disease Classification on PADS (cross-validation)
93.9Accuracy (PD vs DD)SSL Linear Probe
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| SSL Linear ProbeTraining Data Percentage=100%, Evaluation Protocol=Linear Probing2026.04 | 93.9 | 94.12 | 94.01 | — | |
| TimesFM + LoRABackbone=TimesFM, Model Adaptation=LoRA2026.04 | 92.89 | 93.62 | 93.26 | — | |
| Nested CV (Optuna)Optimization Strategy=Nested CV, Optimizer=Optuna2026.04 | 92.57 | 93.61 | 93.09 | — | |
| SSL Fine-tuneTraining Data Percentage=100%, Evaluation Protocol=Fine-tuning2026.04 | 92.5 | 93.56 | 93.03 | — | |
| SSL Fine-tuneTraining Data Percentage=20%, Evaluation Protocol=Fine-tuning2026.04 | 91.67 | 92.26 | 91.96 | — | |
| Three-Class Baseline2026.04 | 87.37 | — | 87.37 | — | |
| Base ModelBack-propagation (BP) setting=w/ BP2026.04 | 87.04 | 93.12 | 90.08 | — | |
| TimesFM Zero-shot LearningBackbone=TimesFM, Evaluation Protocol=Zero-shot Learning2026.04 | 85.42 | 87.39 | 86.41 | — | |
| Base ModelBack-propagation (BP) setting=w/o BP2026.04 | 85.39 | 85.98 | 85.69 | — | |
| Proposed ModelDeployment Hardware=Raspberry Pi2026.04 | — | — | — | 48.32 |