Research map / Architectures and efficiency / Pruning and efficient tuning
What Makes a Good Prune? Maximal Unstructured Pruning for Maximal Cosine Similarity
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a view of unstructured pruning as maximizing cosine similarity between pruned and unpruned network outputs/weights, offering an explanation for pruning behavior and training dynamics.
Problem
Understanding the mechanisms underlying neural network pruning: why methods prune by different percentages yet reach similar performance, why pruning at initialization is hard, and why some models are more prunable.
Approach
Frames maximal unstructured pruning as a way to preserve cosine similarity with the unpruned model, arguing that good prunes maximize this similarity and that this explains pruning behavior.
Experimental setup
Metrics: cosine similarity
Closest papers by meaning
- Rethinking the Value of Network Pruning (ICLR 2019 · similarity 0.92)
- Pruning from Scratch (AAAI 2020 · similarity 0.92)
- How Sparse Can We Prune A Deep Network: A Fundamental Limit Perspective (NeurIPS 2024 · similarity 0.90)
- Targeted Low-rank Refinement: Enhancing Sparse Language Models with Precision (ICML 2025 · similarity 0.90)
- Sparsest Models Elude Pruning: An Exposé of Pruning’s Current Capabilities (ICML 2024 · similarity 0.89)
- Distilling the Knowledge in Data Pruning (ICML 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.