Research Papers 论文研究 4h ago Updated 23m ago 更新于 23分钟前 45

Mixture of Channel Experts: Static Sparse Supports with Input-Adaptive Mixing for Pointwise Projections 通道专家混合:用于逐点投影的静态稀疏支撑与输入自适应混合

Mixture-of-Experts (MoE) design fails when directly applied to convolutional networks because parallel convolutional experts reading the same input channels learn nearly identical filters MoCE shifts the expert axis from operator duplication to channel selection, where each expert is a single output channel with a learned sparse support of k << C input channels The method uses an input-adaptive softmax temperature (predicted per input) enabling dynamic interpolation between mean-like and max-lik 提出Mixture of Channel Experts (MoCE),将MoE思想从算子复制迁移到通道选择,解决卷积网络中并行专家学习同质化滤波器的问题 MoCE每个专家对应单个输出通道,具有学习的稀疏输入通道支持(k << C),通过输入自适应softmax温度实现均值到最大值的动态聚合 引入残差专家汇总未选中通道,配合负载均衡损失确保通道覆盖完整,计算成本从O(C²)降至O(k/C) 在ResNet、EfficientViT等架构上验证,在ImageNet-1K和CIFAR-100上匹配或超越稠密基线,MACs减少16.7%且端到端延迟降低

58
Hot 热度
72
Quality 质量
65
Impact 影响力

Analysis 深度分析

TL;DR

  • Mixture-of-Experts (MoE) design fails when directly applied to convolutional networks because parallel convolutional experts reading the same input channels learn nearly identical filters
  • MoCE shifts the expert axis from operator duplication to channel selection, where each expert is a single output channel with a learned sparse support of k << C input channels
  • The method uses an input-adaptive softmax temperature (predicted per input) enabling dynamic interpolation between mean-like and max-like aggregation
  • MoCE replaces dense pointwise projections with O(k/C) relative cost, achieving 16.7% MAC reduction and lower end-to-end latency while matching or exceeding dense baselines

Why It Matters

This work addresses a fundamental architectural question about how sparse routing paradigms from language models can be meaningfully adapted to vision backbones, offering a principled alternative to dense channel-mixing layers. For practitioners building efficient vision models, MoCE provides a drop-in replacement for 1x1 convolutions that delivers measurable latency and compute savings without accuracy degradation.

Technical Details

  • Core innovation: Each expert in MoCE corresponds to a single output channel with a sparse support over k << C input channels, replacing the dense pointwise (1x1) projection used in standard channel-mixing layers
  • Input-adaptive aggregation: A softmax over selected channels uses a per-input predicted temperature, allowing each expert to dynamically shift between mean-like (high temperature) and max-like (low temperature) aggregation behavior
  • Residual expert: A dedicated residual expert summarizes the unselected channels, ensuring no information is lost from channels outside the sparse support
  • Load-balancing loss: A regularization term enforces complete channel coverage across experts, preventing collapse to a subset of channels
  • Evaluation: Tested across ResNet backbones on ImageNet-1K and CIFAR-100, transfer learning settings, and EfficientViT architectures, achieving 16.7% MAC reduction and improved end-to-end latency while matching or exceeding dense baselines

Industry Insight

  • The failure mode identified—parallel convolutional experts learning identical filters—suggests that naive MoE-to-CNN transfers may require structural modifications beyond simple architectural substitution; channel-level sparsity should be considered a first-class design choice
  • MoCE's input-adaptive temperature mechanism offers a generalizable pattern for dynamic computation in vision transformers and CNNs, potentially applicable beyond channel-mixing layers to other pointwise operations
  • The demonstrated wall-clock savings alongside theoretical MAC reductions indicate that sparse channel selection is practically viable for deployment, making it attractive for edge and mobile vision applications where latency is critical

TL;DR

  • 提出Mixture of Channel Experts (MoCE),将MoE思想从算子复制迁移到通道选择,解决卷积网络中并行专家学习同质化滤波器的问题
  • MoCE每个专家对应单个输出通道,具有学习的稀疏输入通道支持(k << C),通过输入自适应softmax温度实现均值到最大值的动态聚合
  • 引入残差专家汇总未选中通道,配合负载均衡损失确保通道覆盖完整,计算成本从O(C²)降至O(k/C)
  • 在ResNet、EfficientViT等架构上验证,在ImageNet-1K和CIFAR-100上匹配或超越稠密基线,MACs减少16.7%且端到端延迟降低

为什么值得看

本文揭示了MoE架构在视觉卷积网络中直接迁移的结构性缺陷,为高效视觉模型设计提供了新的稀疏混合范式。MoCE通过通道级稀疏选择替代算子级复制,在保持精度的同时显著降低计算开销,对移动端和边缘设备部署具有实用价值。

技术解析

  • 核心创新:将专家轴从"算子复制"转向"通道选择",每个专家定义为单个输出通道+稀疏输入通道支持(k << C),避免并行卷积专家学习同质化滤波器的问题
  • 自适应聚合机制:使用按输入预测温度的softmax组合选中通道,使每个专家能在均值聚合(高温度)和最大值聚合(低温度)之间动态切换
  • 完整性保障:残差专家负责汇总未选中通道的信息,配合负载均衡损失确保所有输入通道被充分覆盖,避免信息丢失
  • 计算效率:稠密投影成本随通道数C二次增长,MoCE相对成本缩放为k/C,实测验证了理论预测的墙钟时间节省
  • 实验验证:在ResNet骨干网络、ImageNet-1K、CIFAR-100分类任务,以及Transfer Learning和EfficientViT场景下测试,均匹配或超越稠密基线和先前通道选择方法

行业启示

  • 视觉模型稀疏化应聚焦特征维度(通道/空间)而非简单复制算子,MoE的成功迁移需要适配视觉数据的结构特性
  • 输入自适应机制(如温度预测)可为模型提供动态计算能力,在精度和效率之间实现更灵活的权衡
  • 通道级稀疏混合为高效视觉 backbone 设计提供了新思路,有望推动移动端和边缘AI应用的模型优化

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

LLM 大模型 Training 训练 Research 科学研究