Files
myWiki/concepts/dynamic-relation-modeling.md

36 lines
1.1 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

---
title: "Dynamic Relation Modeling"
created: 2026-05-15
updated: 2026-05-15
type: concept
tags: [machine-learning, graph-neural-networks, relational-data]
sources: [raw/papers/zeng-neurida-2025.md]
---
# Dynamic Relation Modeling
**Dynamic Relation Modeling** 是 DIME 管线的第二阶段,负责将跨表关系结构融入元组表示。
## 机制
在 [[relational-graph|关系图]](以 FK-PK 为边的元组图)上执行**关系感知消息传递**
1. 每个元组节点以其 [[base-table-embedding|Base Table Embedding]] 初始化
2. 消息沿 FK-PK 边传递,聚合邻接表元组的信息
3. 通过多轮消息传递,元组嵌入吸收关联表的语义信息
4. 输出:**关系嵌入Relational Embedding**,融合了表内语义和跨表结构
## 可配置维度
- 聚合器类型min / max / mean / sum
- 消息传递层数
- 编码器层数和注意力头数
## 消融实验
移除 Dynamic Relation Modeling 后,性能显著下降——证实跨表结构信息对预测至关重要,尤其在目标表特征稀疏的场景下。
## 来源
- [[zeng-neurida-2025|NeurIDA 论文]]