分形盆地几何与深度学习可预测性、可复现性的极限

Nature Machine Intelligence,在线发表:2026年11月5日;doi:10.1038/s42256-026-01336-3
译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。
查看英文原文(Fractal basin geometry and the limits to predictability and reproducibility in deep learning)
Nature Machine Intelligence, Published online: 05 November 2026; doi:10.1038/s42256-026-01336-3
Deep learning outcomes are hard to predict and reproduce. Ly and Gong propose fractal basin geometry as a mechanism: starting points leading to different outcomes interweave at arbitrarily fine scales, so greater precision yields little improvement.正文保留英文原文,标题与摘要由机器翻译。
本文由采集器自动同步
阅读原文 ↗
原文发布于 Nature Machine Intelligence(Fractal basin geometry and the limits to predictability and reproducibility in deep learning),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。