Prior-data fitted networks (PFNs) show that pretraining on large populations of synthetic tasks can enable generalization to unseen datasets with little or no downstream fine-tuning. This paradigm is effective for transformer-based tabular and relational foundation models such as RDB-PFN, but remains largely unexplored for graph-based relational models built on message passing. We pretrain the graph-based relational foundation model Griffin on large collections of synthetic relational databases and ask whether curriculum learning, which benefits transformer-based relational learners, also improves message-passing representations. Rather than fine-tune on downstream tasks, we use the pretrained encoder as a general-purpose relational representation learner and pass its latent representations to a frozen in-context predictor (\ticl{}); no component is trained on the target task. On 16 real-world relational classification tasks, evaluated over five seeds against raw-feature and untrained-encoder baselines, a purely synthetic curriculum that begins with schema-guided data and continues with schema-agnostic data gives the strongest frozen in-context performance\rev{, outperforming real-data-only pretraining and every unordered single-stage mixture of the corpora; when appended after real-data pretraining, it yields the best encoders overall}. These results provide initial evidence that PFN-style synthetic pretraining transfers to graph-based relational foundation models.
Learning Relational Tabular Data without Shared Features
National University of Singapore(新加坡国立大学);Institute of Data Science, National University of Singapore(新加坡国立大学数据科学研究院);Department of Computer Science, National University of Singapore(新加坡国立大学计算机学院);Department of Mathematics, National University of Singapore(新加坡国立大学数学学院)
Is Fixing Schema Graphs Necessary? Full-Resolution Graph Structure Learning for Relational Deep Learning
SKLCCSE, School of Computer Science and Engineering, Beihang University, Beijing, China(信息与电子学院,北京航空航天大学,北京,中国);Key Lab of Education Blockchain and Intelligent Technology, Ministry of Education, Guangxi Normal University, Guilin, China(教育区块链与智能技术重点实验室,教育部,广西师范大学,桂林,中国)
Relational Deep Learning: Graph Representation Learning on Relational Databases
Stanford University(斯坦福大学);Max Planck Institute for Informatics(马克斯·普朗克信息学研究所);Yale University(耶鲁大学);University of Illinois at Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校)
本文提出 Relational Deep Learning,将关系数据库建模为主外键连接的时序异构图,并用消息传递 GNN 端到端学习,配合 RelBench 基准实现多表数据的直接建模。
TablePuppet: A Generic Framework for Relational Federated Learning
ETH Zürich(苏黎世联邦理工学院);UIUC(伊利诺伊大学厄巴纳-香槟分校);Institute of Software, Chinese Academy of Sciences(中国科学院软件研究所);University of Chicago(芝加哥大学);Tsinghua University(清华大学);Microsoft Research(微软研究院)
Relational In-Context Learning via Synthetic Pre-training with Structural Prior
Institute for Artificial Intelligence, Peking University(北京大学人工智能研究院);University of Illinois at Urbana-Champaign(伊利诺伊大学香槟分校);Institute of Computing Technology, Beijing University of Post(北京邮电大学计算机学院);State Key Laboratory of General Artificial Intelligence(通用人工智能国家重点实验室)
School of Data Science(数据科学学院);School of Computer Science and Technology(计算机科学与技术学院);University of Science and Technology of China(中国科学技术大学)
核心集是可以近似地保持原始输入数据集的结构的小集合。因此,我们可以在一个核心集上运行我们的算法,以减少总的计算复杂性。传统的coreset技术假定输入数据集可用于显式处理。然而,这种假设在现实世界中可能不成立。在本文中,我们考虑问题的coresets建设关系数据。也就是说,数据被解耦到几个关系表中,并且通过连接表来直接具体化数据矩阵可能非常昂贵。我们提出了一种新的方法,称为“聚合树与伪立方体”,可以建立一个核心集从下到上。此外,我们的方法可以巧妙地规避关系学习问题的几个麻烦问题[Khamis et al.,PODS 2019]。在一些温和的假设下,我们证明了我们的核心集方法可以应用于机器学习任务,如聚类,逻辑回归和SVM。
Synthesis Without Training: An Inference-Only Pipeline for Tabular, Temporal, and Relational Synthetic Data
Betterdata AI;National University of Singapore(新加坡国立大学);University of Illinois Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校);KAJIMA Technical Research Institute Singapore(鹿岛技术研究所新加坡分所)
AutoSmart: An Efficient and Automatic Machine Learning framework for Temporal Relational Data
DeepBlue Technology(深兰科技);Peking University(北京大学)
时态关系数据可能是工业机器学习应用中最常用的数据类型,需要劳动密集型的特征工程和数据分析来提供精确的模型预测。需要一个自动的机器学习框架来减轻手动微调模型的工作,以便专家可以更多地关注真正需要人类参与的其他问题,例如问题定义,部署和业务服务。然而,时态关系数据的自动化解决方案面临着三个主要挑战:1)如何有效地自动从多个表中挖掘出有用的信息,并从这些表中挖掘出关系。2)如何自我调节,将时间和内存消耗控制在一定的预算内?以及3)如何对广泛的任务给出通用的解决方案?在这项工作中,我们提出了我们的解决方案,成功地解决了上述问题,在一个端到端的自动方式。拟议的框架AutoSmart是AutoML赛道KDD Cup 2019的获胜解决方案,这是迄今为止最大的AutoML比赛之一(860支球队,约4,955份参赛作品)。该框架包括自动数据处理,表合并,特征工程,模型调整,与时间\存储器控制器,有效地和自动制定的模型。拟议的框架在各个领域的多个数据集上显着优于基线解决方案。
4DBInfer: A 4D Benchmarking Toolbox for Graph-Centric Predictive Modeling on Relational DBs
Amazon Web Services(亚马逊网络服务);Shanghai Jiaotong University(上海交通大学);The Hong Kong University of Science and Technology(香港科技大学);Fudan University(复旦大学);Tsinghua University(清华大学);Southern University of Science and Technology(南方科技大学);Peking University(北京大学);The Hong Kong Polytechnic University(香港理工大学);Carnegie Mellon University(卡内基梅隆大学)