Controllable Data Generation by Deep Learning: A Review
Shiyu Wang Department of Biostatistics and Bioinformatics, Emory UniversityUSA , Yuanqi Du Department of Computer Science, Cornell UniversityUSA , Xiaojie Guo JD.COM Silicon Valley Research CenterUSA , Bo Pan Department of Computer Science, Emory UniversityUSA , Zhaohui Qin Department of Biostatistics and Bioinformatics, Emory UniversityUSA and Liang Zhao Department of Computer Science, Emory UniversityUSA
Abstract
Designing and generating new data under targeted properties has been attracting various critical applications such as molecule design, image editing and speech synthesis. Traditional hand-crafted approaches heavily rely on expertise experience and intensive human efforts, yet still suffer from the insufficiency of scientific knowledge and low throughput to support effective and efficient data generation. Recently, the advancement of deep learning has created the opportunity for expressive methods to learn the underlying representation and properties of data. Such capability provides new ways of determining the mutual relationship between the structural patterns and functional properties of the data and leveraging such relationships to generate structural data, given the desired properties. This article is a systematic review that explains this promising research area, commonly known as controllable deep data generation. First, the article raises the potential challenges and provides preliminaries. Then the article formally defines controllable deep data generation, proposes a taxonomy on various techniques and summarizes the evaluation metrics in this specific domain. After that, t
中文速览
如何让深度学习模型按照人们的需求"定向"生成数据，是分子设计、图像编辑、语音合成等领域共同面临的核心难题——光靠传统手工方法既依赖稀缺的领域专家知识，又难以应对动辄高达10³³量级的巨大搜索空间。这篇综述系统梳理了"可控深度数据生成"（controllable deep data generation）这一新兴方向，从问题定义出发，提出了一套统一的方法分类体系，覆盖分子优化、蛋白质设计、图像编辑、情感语音合成等主要应用场景，并整理了跨领域通用的评估指标与基准数据集。研究发现，尽管已有大量方法在各自领域取得突破，但多属性协同控制、离散结构的高效表示以及属性标注成本高等挑战仍制约着整体进展。这项工作的价值在于为来自不同领域的研究者提供了一张"导航图"，帮助他们快速定位适合自身问题的技术路线，同时也为推动各应用领域之间的方法迁移与交叉融合奠定了基础。
原文 arXiv:2207.09542;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2207.09542v6