English(EN) WBMM: Windowed Batch Matrix Multiplication for Efficient Large Receptive Field Convolution

新的WBMM技术提高了大卷积核的效率

作者 PulseAugur 编辑部 · [2 个来源] · 2026-07-02 12:33

研究人员开发了一种名为窗口化批处理矩阵乘法（WBMM）的新技术，以提高大卷积核深度卷积的效率。传统方法随着卷积核尺寸的增加而出现性能下降，但WBMM将输入分割成窗口并使用偏置表构建权重矩阵，通过批处理矩阵乘法实现规则的内存访问。该方法在更大的窗口下显示出更高的吞吐量，并在ImageNet-1K、COCO和ADE20K等基准测试中取得了相当或更好的准确率，同时在各种硬件平台上实现了显著的训练加速。 AI

影响 WBMM为需要大感受野的模型提供了更高效的训练和推理途径，有可能在各种硬件上提高性能。

排序理由详细介绍深度学习卷积新计算技术的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。我们如何撰写摘要 →

报道来源 [2]

arXiv cs.LG TIER_1 English(EN) · Wan Song, Wei Zhou, Rui Wang, Jun Yu, Toru Kurihara, Jiajia Xu, Shu Zhan · 2026-07-03 04:00

WBMM: Windowed Batch Matrix Multiplication for Efficient Large Receptive Field Convolution

arXiv:2607.02097v1 Announce Type: cross Abstract: Large kernel depthwise convolutions achieve strong performance but suffer from significant degradation as kernel size grows due to irregular memory access from gather-based computation; while Large Kernel Acceleration (LKA) helps …
arXiv cs.LG TIER_1 English(EN) · Shu Zhan · 2026-07-02 12:33

WBMM: Windowed Batch Matrix Multiplication for Efficient Large Receptive Field Convolution

Large kernel depthwise convolutions achieve strong performance but suffer from significant degradation as kernel size grows due to irregular memory access from gather-based computation; while Large Kernel Acceleration (LKA) helps on small feature maps, it becomes counterproductiv…

报道来源 [2]

WBMM: Windowed Batch Matrix Multiplication for Efficient Large Receptive Field Convolution

WBMM: Windowed Batch Matrix Multiplication for Efficient Large Receptive Field Convolution

相关实体

相关话题