将输入分解为可解释的多个语义成分，实现更清晰的特征表达。
Decomposer Networks: Deep Component Analysis and Synthesis
- 通过并行分支与互减残差机制，实现成分间显式竞争
- 每个成分独立重建，整体保持语义可解释性与压缩效率
- 适合需要可解释性分解的图像/信号建模任务
我们提出分解网络（DecompNet），一种语义自编码器，能将输入信号分解为多个可解释的组成部分。不同于传统自编码器将输入压缩为单一潜在表示，DecompNet采用N条并行分支，每条分支的输入定义为原始信号减去其他所有分支的重建结果。通过将高斯-赛德尔风格的块坐标下降算法展开为可微网络，DecompNet强制各成分间显式竞争，从而生成简洁且具有语义意义的表示。我们在理论上将模型与线性分解方法（如PCA、NMF）、深度展开优化及物体中心架构（MONet、IODINE、Slot Attention）进行对比，强调其作为首个实现‘除一个外其余全减’残差更新规则的语义自编码器的独特性。
原文摘要 · Abstract (English)
We propose the Decomposer Networks (DecompNet), a semantic autoencoder that factorizes an input into multiple interpretable components. Unlike classical autoencoders that compress an input into a single latent representation, the Decomposer Network maintains N parallel branches, each assigned a residual input defined as the original signal minus the reconstructions of all other branches. By unrolling a Gauss--Seidel style block-coordinate descent into a differentiable network, DecompNet enforce explicit competition among components, yielding parsimonious, semantically meaningful representations. We situate our model relative to linear decomposition methods (PCA, NMF), deep unrolled optimization, and object-centric architectures (MONet, IODINE, Slot Attention), and highlight its novelty as the first semantic autoencoder to implement an all-but-one residual update rule.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。