LightGBM
PulseAugur coverage of LightGBM — every cluster mentioning LightGBM across labs, papers, and developer communities, ranked by signal.
6 天有情绪数据
-
Foundation models show promise in disease prediction and RF loss classification
Researchers have evaluated the Tabular Pre-Trained Foundation Network (TabPFN) for predicting the conversion of Mild Cognitive Impairment to Alzheimer's Disease, finding it outperforms traditional machine learning model…
-
用于可解释、公平和可观察医院再入院预测的集成框架：在 MIMIC-IV 上的开发与验证
研究人员开发了一种新的梯度正则化牛顿方案，以确保梯度提升决策树 (GBDT) 的全局收敛性，这是一种广泛用于表格机器学习的技术。该方法引入了一个自适应 L2 正则化项，实现了与 Nesterov 动量等一阶提升方法相当的收敛速度。数值实验表明，该新方案在标准牛顿提升可能发散的地方也能收敛。此外，另一项研究提出了一个用于从心电图中诊断射血分数的模态机器学习框架，实现了高精度并提供了可解释的特征。
-
Japanese medical foundation model shows task-dependent optimal scale
Researchers have investigated the relationship between model scale and performance for structured medical foundation models using a large Japanese claims database. Their findings indicate that optimal model size varies …