提出新剪枝方法，让模型在高压压缩下仍保持高性能。
IPPRO: Importance-based Pruning with PRojective Offset for Magnitude-indifferent Structural Pruning
- 用投影几何重构滤波器重要性，避免尺度干扰
- 单步梯度就能预测多步剪枝效果，效率更高
- 适合作为通用剪枝框架，适合追求高压缩率的场景
基于重要性的结构化剪枝普遍依赖滤波器幅度，但该代理存在根本缺陷：由于尺度不变性，功能相同的滤波器在缩放后可能获得任意不同的重要性评分。本文提出IPPROM（基于重要性的剪枝方法，结合投影偏移），一种基于射影几何的尺度不变剪枝框架。通过将滤波器嵌入实射影空间（$b{RP}^N$），IPPROM解决了原点奇点问题，使所有滤波器与零滤波器保持等角距离。我们定义了PROscore，通过测量单次梯度步长下滤波器向零方向的角位移（方向坍缩）来捕捉功能重要性。进一步证明PROscore与精确的$L_0$松弛相关，该一次性准则可可靠预测多步剪枝动态。在CNN、视觉变换器和大语言模型（如ResNet、DeiT、LLaMA）上的大量实验表明，IPPROM持续优于现有方法，在高压缩率且无需微调的情况下表现尤为突出，建立了鲁棒、架构无关的神经网络压缩范式。
原文摘要 · Abstract (English)
Importance-based structured pruning overwhelmingly relies on filter magnitude. This proxy is fundamentally flawed: due to scale invariance, functionally identical filters can receive arbitrarily different importance scores under rescaling. We propose IPPRO (Importance-based Pruning with PROjective Offset), a scale-invariant pruning framework grounded in projective geometry. By embedding filters into real projective space ($\mathbb{RP}^N$), IPPRO resolves the singularity at the origin, placing all filters at an equal angular distance from the zero filter. We define PROscore, which captures functional importance by measuring a filter's angular displacement toward zero under a single gradient step (directional collapse). We further connect PROscore to exact $L_0$ relaxation, proving this one-shot criterion reliably predicts multi-step pruning dynamics. Extensive experiments across CNNs, Vision Transformers, and LLMs (e.g., ResNet, DeiT, LLaMA) demonstrate that IPPRO consistently outperforms existing methods, yielding particularly striking gains under high compression and no-fine-tuning regimes, IPPRO establishes a robust, architecture-agnostic paradigm for neural network compression.
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。