L2T-DFM: Learning to Teach with Dynamic Fused Metric

Open Webpage

Zhaoyang Hai, Liyuan Pan, Xiabi Liu, Mengqiao Han

Published: 2025, Last Modified: 09 Nov 2025Pattern Recognit. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Propose a DFM-based teaching approach to generate loss function dynamically for the neural network optimization.•Propose a confidence-based selection algorithm to select appropriate metrics.•Employ information divergence to assist in integrating the states of a student model to enhance the teacher model.•Conduct extensive experiments on a wide range of loss functions and tasks to demonstrate the effectiveness of our approach.