PulseAugur
中
实时 18:21:30

Reducing Per-Sample Harm in Stochastic Optimization

作者 PulseAugur 编辑部 · [1 个来源] ·

研究人员开发了一种新方法来减少 AI

排序理由 [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 →

Reducing Per-Sample Harm in Stochastic Optimization

报道来源 [1]

  1. arXiv stat.ML TIER_1 English(EN) · Apostolos Avranas ·

    降低随机优化中每样本的损害

    arXiv:2607.16261v1 Announce Type: cross Abstract: Modern optimizers combine gradients from the current mini-batch with historical optimization state, such as momentum or adaptive moments. While highly effective, aggregating across the batch and incorporating this history can prod…