Approximate Query Processing using Deep Generative Models
This addresses the need for faster data exploration and visualization in interactive applications, representing an incremental improvement by applying deep learning to an existing database technique.
The paper tackles the problem of slow data analysis by using deep generative models for Approximate Query Processing (AQP) to answer aggregate queries quickly, achieving high accuracy and low latency with compact models of a few hundred KBs.
Data is generated at an unprecedented rate surpassing our ability to analyze them. The database community has pioneered many novel techniques for Approximate Query Processing (AQP) that could give approximate results in a fraction of time needed for computing exact results. In this work, we explore the usage of deep learning (DL) for answering aggregate queries specifically for interactive applications such as data exploration and visualization. We use deep generative models, an unsupervised learning based approach, to learn the data distribution faithfully such that aggregate queries could be answered approximately by generating samples from the learned model. The model is often compact - few hundred KBs - so that arbitrary AQP queries could be answered on the client side without contacting the database server. Our other contributions include identifying model bias and minimizing it through a rejection sampling based approach and an algorithm to build model ensembles for AQP for improved accuracy. Our extensive experiments show that our proposed approach can provide answers with high accuracy and low latency.