Unverified50% confidenceFactExact time
决策树的分裂准则（信息增益、基尼不纯度、均方误差）本质上是排序统计量的函数，只关心阈值两侧的类别比例而非样本间距离
1
Sources
50%
Confidence
Long-term
Relevance
8/24/2026
First Seen
Sources
Related Entities
Related Claims
Unverified决策树方法通过一系列关键问题（任务时间跨度、信息规模与检索需求、成本与延迟约束）逐步收敛到最合适的记忆方案67% similarUnverified路由分类器需同时满足轻量（推理耗时毫秒级）和准确两个要求，误判会导致质量下降或资源浪费65% similarUnverified多个独立模型的分歧程度本质上是对数据标注难度的一种估计，高分歧样本往往对应边界案例或存在真实争议的问题64% similarUnverifiedThe choice of bottleneck dimensionality is essentially a trade-off between information compression rate and reconstruction quality, corresponding to rate-distortion theory; Shannon proved in 1959 that there exists a theoretical minimum coding rate lower bound for a given distortion tolerance.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/793941API
curl https://kongchang.com/api/v1/knowledge/claims/793941MCP
get_claim(id=793941)