开源HPC系统ExaTune及超参数景观基准发布

1. 13:00
   arXiv 机器学习

   ExaTune：大规模穷举超参数景观探索

   开源 HPC 系统 ExaTune 推出，通过穷举评估超参数网格构建超参数景观图，并引入五项形式化景观指标。团队在 HuggingFace 发布包含 8 个景观图、共 285,080 个配置的基准集，涵盖 XGBoost 等 5 种模型族。实证分析表明，Optuna 与 Hyperband 等方法在标准预算下会系统性遗漏不稳定性区域及参数交互效应。

Current heat 6·Comparable peak 10（Nov 4 14:00）·Comparable change, last 24 hours–

Trends compare only the same parties observed continuously, so the scope can be narrower than current heat counts. Move the pointer or tap the chart to see hourly heat; use the left and right arrow keys on a keyboard.