通过融合特征图收敛评估，提升模型性能且无需修改网络结构。
FMCE-Net++: Feature Map Convergence Evaluation and Training
- 引入预训练的FMCE-Net作为辅助头，生成特征收敛评分指导训练。
- 在多个数据集上实现1.08~1.16个百分点的准确率提升。
- 适合关注模型可解释性与性能优化的研究者使用。
深度神经网络因内部表示不透明而面临可解释性挑战。尽管特征图收敛评估（FMCE）可通过特征图收敛评分（FMCS）量化模块级收敛程度，但缺乏实验验证和闭环整合。为此，本文提出FMCE-Net++，一种新型训练框架，将预训练、冻结的FMCE-Net作为辅助头，生成FMCS预测值，结合任务标签，通过表示辅助损失（RAL）联合监督主干网络优化。RAL通过可调的‘表示抽象因子’动态平衡主分类损失与特征收敛优化。在MNIST、CIFAR-10、FashionMNIST和CIFAR-100上的大量实验表明，FMCE-Net++在不改变网络结构或增加数据的前提下，持续提升模型性能。关键结果包括：在ResNet-50/CIFAR-10上提升+1.16个百分点，在ShuffleNet v2/CIFAR-100上提升+1.08个百分点，验证了该方法能有效突破现有性能上限。
原文摘要 · Abstract (English)
Deep Neural Networks (DNNs) face interpretability challenges due to their opaque internal representations. While Feature Map Convergence Evaluation (FMCE) quantifies module-level convergence via Feature Map Convergence Scores (FMCS), it lacks experimental validation and closed-loop integration. To address this limitation, we propose FMCE-Net++, a novel training framework that integrates a pretrained, frozen FMCE-Net as an auxiliary head. This module generates FMCS predictions, which, combined with task labels, jointly supervise backbone optimization through a Representation Auxiliary Loss. The RAL dynamically balances the primary classification loss and feature convergence optimization via a tunable \Representation Abstraction Factor. Extensive experiments conducted on MNIST, CIFAR-10, FashionMNIST, and CIFAR-100 demonstrate that FMCE-Net++ consistently enhances model performance without architectural modifications or additional data. Key experimental outcomes include accuracy gains of $+1.16$ pp (ResNet-50/CIFAR-10) and $+1.08$ pp (ShuffleNet v2/CIFAR-100), validating that FMCE-Net++ can effectively elevate state-of-the-art performance ceilings.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。