arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2208.07463cs.CVcs.AI
收藏

Conv-Adapter:探索卷积神经网络的参数高效迁移学习

Conv-Adapter: Exploring Parameter Efficient Transfer Learning for ConvNets

Peking University(北京大学);Microsoft Research Asia(微软亚洲研究院)

Hao Chen, Ran Tao, Han Zhang, Yidong Wang, Xiang Li, Wei Ye, Jindong Wang, Guosheng Hu, Marios Savvides

AI总结:

针对参数高效调优方法在大规模卷积神经网络视觉任务中研究不足的问题,提出轻量可迁移的Conv-Adapter模块,仅用少量参数就在多类视觉任务上达到或超越全微调性能。

摘要:

尽管参数高效调优(PET)方法在自然语言处理(NLP)任务的Transformer架构中已展现出巨大潜力,但它们在计算机视觉(CV)任务中针对大规模卷积神经网络(ConvNets)的有效性仍有待充分研究。本文提出了Conv-Adapter,一种专为ConvNets设计的PET模块。Conv-Adapter轻量级、可跨领域迁移且架构无关,在不同任务上具备泛化性能。在迁移到下游任务时,Conv-Adapter在冻结预训练参数的同时,学习对骨干网络中间表示的任务特定特征调制。它仅引入极少量可学习参数,例如仅为ResNet50全微调参数的3.5%,也可应用于基于Transformer的骨干网络。Conv-Adapter在23个不同领域的分类任务上优于此前的PET基线方法,性能与全微调相当甚至更优。它在少样本分类中也表现出优越性能,平均领先幅度为3.39%。除分类任务外,Conv-Adapter可泛化到检测和分割任务,参数减少超过50%但性能与传统全微调相当。

英文摘要:

While parameter efficient tuning (PET) methods have shown great potential with transformer architecture on Natural Language Processing (NLP) tasks, their effectiveness with large-scale ConvNets is still under-studied on Computer Vision (CV) tasks. This paper proposes Conv-Adapter, a PET module designed for ConvNets. Conv-Adapter is light-weight, domain-transferable, and architecture-agnostic with generalized performance on different tasks. When transferring on downstream tasks, Conv-Adapter learns tasks-specific feature modulation to the intermediate representations of backbones while keeping the pre-trained parameters frozen. By introducing only a tiny amount of learnable parameters, e.g., only 3.5% full fine-tuning parameters of ResNet50. It can also be applied for transformer-based backbones. Conv-Adapter outperforms previous PET baseline methods and achieves comparable or surpasses the performance of full fine-tuning on 23 classification tasks of various domains. It also presents superior performance on the few-shot classification with an average margin of 3.39%. Beyond classification, Conv-Adapter can generalize to detection and segmentation tasks with more than 50% reduction of parameters but comparable performance to the traditional full fine-tuning.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑