Compact deep models are attractive for medical image classification, but conventional knowledge distillation mainly transfers class-level predictions and may not adequately preserve lesion-relevant spatial cues. To address this limitation, we propose LaHKD, a lesion-aware hierarchical knowledge distillation framework f...
Ya-Rong Liu, Run-Mei Xie, Xiao-Lan Xie et al.· Journal of Imaging· 0 citations
KCD-GEIRTimeXer is proposed, a short-term electric load forecasting method built upon TimeXer that reduces MSE, RMSE, MAE, and MAPE by 26.25% and outperforms all baseline models, respectively.
The proposed MST-iTransformer model achieves simultaneous reductions in prediction errors and inference latency, validating the effectiveness of the multi-scale temporal encoding, sparse attention mechanism, and adaptive feature selection modules in improving forecasting accuracy and computational efficiency.