惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

月光博客
月光博客
人人都是产品经理
人人都是产品经理
博客园 - 聂微东
WordPress大学
WordPress大学
S
SegmentFault 最新的问题
博客园 - Franky
V
V2EX
Y
Y Combinator Blog
Google DeepMind News
Google DeepMind News
J
Java Code Geeks
T
The Blog of Author Tim Ferriss
罗磊的独立博客
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
Jina AI
Jina AI
博客园 - 叶小钗
F
Fortinet All Blogs
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
A
About on SuperTechFans
M
MIT News - Artificial intelligence
云风的 BLOG
云风的 BLOG
Last Week in AI
Last Week in AI
D
Docker
博客园 - 【当耐特】
阮一峰的网络日志
阮一峰的网络日志

cs.DB updates on arXiv.org

Block-Sphere Vector Quantization GroupAffect-4: A Multimodal Dataset of Four-Person Collaborative Interaction CogScale: Scalable Benchmark for Sequence Processing TextAlign: Preference Alignment for Text Rendering with Hierarchical Rewards LogRouter: Adaptive Two-Level LLM Routing for Log Question Answering in Big Data Systems Agentic Cost-Aware Query Planning with Knowledge Distillation for Big Data Analytics Covariance Structure and Coordinate Heterogeneity Govern Binary Quantization of Contrastive Embeddings IVF-TQ: Calibration-Free Streaming Vector Search via a Codebook-Free Residual Layer Automatic Unsupervised Ensemble Outlier Model Selection--Extended Version A Generative AI Framework for Intelligent Utility Billing CO 2 Analytics and Sustainable Resource Optimisation Towards Foundation Models for Relational Databases with Language Models and Graph Neural Networks Gaussian Relational Graph Transformer Croissant Baker: Metadata Generation for Discoverable, Governable, and Reusable ML Datasets Reducing Hallucination in Vision-Language Models via Stage-wise Preference Optimization under Distribution Shift A Horn extension of DL-Lite with NL data complexity 3D Primitives are a Spatial Language for VLMs Enabling AI-Native Mobility in 6G: A Real-World Dataset for Handover, Beam Management, and Timing Advance A CAP-like Trilemma for Large Language Models: Correctness, Non-bias, and Utility under Semantic Underdetermination EpiCastBench: Datasets and Benchmarks for Multivariate Epidemic Forecasting FERMI: Exploiting Relations for Membership Inference Against Tabular Diffusion Models Toward Multi-Database Query Reasoning for Text2Cypher Autonomous FAIR Digital Objects: From Passive Assertions to Active Knowledge HOME-KGQA: A Benchmark Dataset for Multimodal Knowledge Graph Question Answering on Household Daily Activities Detect, Localize, and Explain: Interactive Hierarchical Log Anomaly Analytics with LLM Augmentation Open Ontologies: Tool-Augmented Ontology Engineering with Stable Matching Alignment Machine Learning-Based Pre-Test Risk Stratification for PCR-Confirmed Chlamydia Using Patient-Reported Data and Urine Biomarkers Reconciling Consistency-Based Diagnosis with Actual-Causality-Based Explanations PrepBench: How Far Are We from Natural-Language-Driven Data Preparation? Anatomy of a Query: W5H Dimensions and FAR Patterns for Text-to-SQL Evaluation Building informative materials datasets beyond targeted objectives
Calibration: A Simple Trick for Wide-table Delta Analytics
Zezhou Huang, Eugene Wu · 2022-10-08 · via cs.DB updates on arXiv.org

Data analytics over normalized databases typically requires computing and materializing expensive joins (wide-tables). Factorized query execution models execution as message passing between relations in the join graph and pushes aggregations through joins to reduce intermediate result sizes. Although this accelerates query execution, it only optimizes a single wide-table query. In contrast, wide-table analytics is usually interactive and users want to apply delta to the initial query structure. For instance, users want to slice, dice and drill-down dimensions, update part of the tables and join with new tables for enrichment. Such Wide-table Delta Analytics offers novel work-sharing opportunities. This work shows that carefully materializing messages during query execution can accelerate Wide-table Delta Analytics by >10^5x as compared to factorized execution, and only incurs a constant factor overhead. The key challenge is that messages are sensitive to the message passing ordering. To address this challenge, we borrow the concept of calibration in probabilistic graphical models to materialize sufficient messages to support any ordering. We manifest these ideas in the novel Calibrated Junction Hypertree (CJT) data structure, which is fast to build, aggressively re-uses messages to accelerate future queries, and is incrementally maintainable under updates. We further show how CJTs benefit applications such as OLAP, query explanation, streaming data, and data augmentation for ML. Our experiments evaluate three versions of the CJT that run in a single-threaded custom engine, on cloud DBs, and in Pandas, and show 30x - 10^5x improvements over state-of-the-art factorized execution algorithms on the above applications.