决策导向学习中识别出雅可比矩阵秩崩溃

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-02 04:00

研究人员在决策导向学习（DFL）中识别出一种称为雅可比矩阵秩崩溃的现象，其中预测器的雅可比矩阵表现出低秩。这种情况意味着优化器使用的梯度是共线的，限制了学习可用的独立参数更新方向。跨越各种配置（包括公平性、最短路径和背包问题）的实验表明，虽然 DFL 可以提供决策质量的适度改进，但这些收益通常是边际的，并且可能对坐标缩放和预测准确性与决策质量等因素敏感。 AI

影响识别出决策导向学习中的一个几何限制，这可能会影响其实际效益。

排序理由关于机器学习中一个新概念的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 13 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 关于机器学习中一个新概念的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, model release; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Aojie Yuan, Haiyue Zhang, Zijian Su · 2026-10-02 04:00

Jacobian Rank Collapse in Decision-Focused Learning

arXiv:2609.39261v1 Announce Type: new Abstract: Decision-focused learning (DFL) trains predictors through downstream objectives, but a different loss need not provide an independent parameter-update direction. We characterize this restriction through the predictor Jacobian, using…

本文如何被排名

报道来源 [1]

Jacobian Rank Collapse in Decision-Focused Learning

相关实体

相关话题