https://www.xin800.com/tags/%E4%BE%9B%E5%BA%94%E9%93%BE/Recent content in 供应链 on 标准答案Hugozh-CNFri, 31 Jul 2026 00:00:00 +0000https://www.xin800.com/post/transformer-attention-pipeline/Fri, 31 Jul 2026 00:00:00 +0000https://www.xin800.com/post/transformer-attention-pipeline/从向量到QKV到点积到缩放到Softmax到加权融合，完整的Transformer注意力流程，全程业务语言，供应链案例贯穿。https://www.xin800.com/post/what-is-cross-attention/Fri, 31 Jul 2026 00:00:00 +0000https://www.xin800.com/post/what-is-cross-attention/交叉注意力机制实现两组不同业务信息之间的关联匹配，Q来自A数据集，K/V来自B数据集，是机器翻译和知识库匹配的核心技术。https://www.xin800.com/post/what-is-self-attention/Fri, 31 Jul 2026 00:00:00 +0000https://www.xin800.com/post/what-is-self-attention/自注意力机制实现同一组数据内部所有条目的两两关联计算，依靠QKV向量与相似度打分自动识别远距离业务关联，是Transformer的核心基础。https://www.xin800.com/post/how-to-judge-vector-proximity/Fri, 31 Jul 2026 00:00:00 +0000https://www.xin800.com/post/how-to-judge-vector-proximity/详解向量相似度判断的底层逻辑：向量不是逐字编码，而是整体语义的打包。通过余弦相似度衡量方向重合度。https://www.xin800.com/post/euclidean-distance-logistics/Fri, 31 Jul 2026 00:00:00 +0000https://www.xin800.com/post/euclidean-distance-logistics/物流网点片区划分、配送路线聚合场景中，欧氏距离不只判断地理远近，还能兼顾货量、时效等量化指标。https://www.xin800.com/post/euclidean-distance-use-cases/Fri, 31 Jul 2026 00:00:00 +0000https://www.xin800.com/post/euclidean-distance-use-cases/欧氏距离同时考量特征方向和数值大小，适合结构化数值聚类、备件指标分类、坐标点位匹配等场景。https://www.xin800.com/post/dot-product-business/Fri, 31 Jul 2026 00:00:00 +0000https://www.xin800.com/post/dot-product-business/点积是一套快速计算公式，用来衡量两组向量大方向是否一致，输出一个数字代表匹配程度，是注意力机制计算关联程度的基础运算。