MemFlow: A Lightweight Forward Memorizing Framework for Quick Domain Adaptive Feature Mapping
Jianming Lv, Chengjun Wang, Depin Liang, Qianli Ma, Wei Chen, Xueqi Cheng
Abstract
Deploying pretrained visual models in real-world environments often suffers from significant performance degradation due to the diversity of testing scenarios. Continuous adaptation of learning models on edge devices via unlabeled data collected from the target domain is highly effective for boosting generalization capability. However, gradient-backpropagation-based optimization of the massive parameters in deep neural networks is vastly more time-consuming than forward inference, rendering online learning infeasible on low-power edge devices. To address this critical challenge, we propose a lightweight gradientfree forward-memorizing framework, namely MemFlow, which leverages a frozen backbone and enables efficient fine-tuning of the mapping between features and predictions. Specifically, MemFlow employs randomly connected neurons to memorize feature-label associations; within the network, spiking signals are propagated, and predictions are generated by associating neuron-stored memories according to their confidence levels. More notably, MemFlow supports reinforced memorization of feature mappings using unlabeled data, thereby enabling rapid adaptation to new domains. Extensive experiments on four real-world cross-domain datasets demonstrate that MemFlow achieves performance improvements of up to 10% while consuming less than 1% of the computational time required by traditional domain adaptation methods. The code is available at https://github.com/so-link/MemFlow
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation cites Larger Norm More Transferable: An Adaptive Feature Norm Approach for Unsupervised Domain Adaptation
- Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation cites Moment Matching for Multi-Source Domain Adaptation
- Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation cites Cluster Alignment With a Teacher for Unsupervised Domain Adaptation
- In Search of Lost Domain Generalization cites Moment Matching for Multi-Source Domain Adaptation
Builds on11
- Moment Matching for Multi-Source Domain AdaptationXingchao Peng, Qinxun Bai, Xide Xia, Zijun Huang et al.ICCV 2019 · 2,239 citations
- Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain AdaptationJian Liang, Dapeng Hu, Jiashi FengICML 2020 · 1,624 citations
- In Search of Lost Domain GeneralizationIshaan Gulrajani, David Lopez-PazICLR 2021 · 1,416 citations
- Larger Norm More Transferable: An Adaptive Feature Norm Approach for Unsupervised Domain AdaptationRuijia Xu, Guanbin Li, Jihan Yang, Liang LinICCV 2019 · 563 citations
- Cluster Alignment With a Teacher for Unsupervised Domain AdaptationZhijie Deng, Yucen Luo, Jun ZhuICCV 2019 · 241 citations
Related papers
- SF-Adapter: Computational-Efficient Source-Free Domain Adaptation for Human Activity RecognitionHua Kang, Qingyong Hu, Qian ZhangUbiComp 2024 · 11 citations
- Curvature-Aware Zeroth-Order Optimization for Memory-Efficient Test-Time AdaptationJunming Zhang, Shuyu Yin, Peilin Liu, Rendong Ying et al.CVPR 2026
- EcoTTA: Memory-Efficient Continual Test-Time Adaptation via Self-Distilled RegularizationJunha Song, Jungsoo Lee, In So Kweon, Sungha ChoiCVPR 2023
- FOZO: Forward-Only Zeroth-Order Prompt Optimization for Test-Time AdaptationXingyu Wang, Tao WangCVPR 2026 · 2 citations
- Test-Time Model Adaptation with Only Forward PassesShuaicheng Niu, Chunyan Miao, Guohao Chen, Pengcheng Wu et al.ICML 2024 · 77 citations
