Research map / Architectures and efficiency / Pruning and efficient tuning
SCHTs: A Semi-Structured Dynamic Sparse Training Framework for Hardware-Efficient Deep Learning
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 SCHTs, a semi-structured dynamic sparse training framework aiming to combine the accuracy of unstructured DST with the hardware efficiency of semi-structured sparsity for edge deployment.
Problem
Unstructured dynamic sparse training yields irregular weight distributions that add index overhead and hurt computational efficiency on edge hardware, while semi-structured sparsity is efficient but typically loses accuracy at high sparsity.
Approach
A semi-structured dynamic sparse training framework (SCHTs) that trains sparse networks with hardware-friendly semi-structured patterns during dynamic topology updates, to close the accuracy gap with unstructured DST. Details are not available beyond the abstract.
Closest papers by meaning
- Dynamic Sparse Training with Structured Sparsity (ICLR 2024 · similarity 0.91)
- Advancing Dynamic Sparse Training by Exploring Optimization Opportunities (ICML 2024 · similarity 0.89)
- A Single-Step, Sharpness-Aware Minimization is All You Need to Achieve Efficient and Accurate Sparse Training (NeurIPS 2024 · similarity 0.89)
- Dynamic Sparse No Training: Training-Free Fine-tuning for Sparse LLMs (ICLR 2024 · similarity 0.88)
- Random Sparse Lifts: Construction, Analysis and Convergence of finite sparse networks (ICLR 2024 · similarity 0.88)
- Dynamic Sparse Training of Diagonally Sparse Networks (ICML 2025 · similarity 0.87)
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.