Skip to content
Hot eventLive

MCL: Meta Convolution Layer 论文发布

1 reports1 sources3 hr ago updated

Get the story

AI overview

2026年10月9日,arXiv 计算机视觉方向发布论文《MCL: Meta Convolution Layer》。论文提出 Meta Convolution Layer(MCL),用高阶多项式展开直接把卷积核建模为输入条件函数 W(x),以嵌套残差块构成结构化多项式元网络,生成单一输入自适应卷积核。该方法摆脱了线性混合动态卷积对基核数量的依赖,并缓解训练不稳定问题。目前公开信息仅涉及该论文本身,未见后续实验验证或第三方复现报道。

Generated from reports · updated 3 hr ago

Timeline

Follow the coverage from different angles.

Oct 9, 2026
  1. arXiv · Computer Vision
    MCL: Meta Convolution Layer

    论文提出 Meta Convolution Layer(MCL),用高阶多项式展开直接把卷积核建模为输入条件函数 W(x),以嵌套残差块构成结构化多项式元网络,生成单一输入自适应卷积核,摆脱线性混合动态卷积对基核数量的依赖并缓解训练不稳定。

Heat trend

Current heat 9·Comparable peak 10(Oct 9)·Comparable change over 24 hours –

02.557.510Oct9Oct9Oct9Oct9

The trend compares only the same participants observed continuously; its range may be smaller than the current heat count. Move or click on the chart to inspect hourly heat; use the left and right arrow keys to switch.