Deutsch(DE) Beyond Wasserstein Proximity: Finding Better Transfer Sources

新的机器学习迁移学习方法优先考虑遥远的数据源

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-27 04:00

研究人员开发了一种新的机器学习模型预训练源选择方法，超越了简单的数据相似性。该方法受领域自适应理论的启发，表明即使增加了Wasserstein距离，更广泛的源变化也能改善估计。使用加密货币波动性数据的实验表明，与最近的源相比，选择最远的Gaussian-Wasserstein源可以将MLP的技能提高12%以上，将Transformer的技能提高20%以上。进一步的分析表明，前向先导选择和回顾性筛选也带来了收益，但运行时节省尚未测量。 AI

影响在迁移学习中引入了一种新颖的源选择方法，通过利用更多样化的数据集来潜在地提高模型性能。

排序理由详细介绍新机器学习方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv stat.ML 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: 22 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 详细介绍新机器学习方法的学术论文。[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, other; 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 stat.ML TIER_1 Deutsch(DE) · Gordei Verbii, Georgii Killinkarov, Vladimir Yafarov · 2026-10-27 04:00

超越Wasserstein距离：寻找更好的迁移源

arXiv:2610.35314v1 Announce Type: new Abstract: Source selection often favors similar data, yet marginal proximity need not identify the most useful pretraining source. We motivate a conditional alternative through domain-adaptation theory and a regression example in which broade…

本文如何被排名

报道来源 [1]

超越Wasserstein距离：寻找更好的迁移源

相关实体

相关话题