通过图像字幕实现差分隐私表示学习
Differentially Private Representation Learning via Image Captioning
AI总结：
针对差分隐私表示学习隐私-准确率权衡不佳的问题，本研究通过图像字幕结合工程技巧，在LAION-2B子集上训练DP-Cap模型，获得高质量特征，在ε=8时ImageNet-1K线性分类准确率达65.8%，显著超越此前最优水平。
摘要：
差分隐私（DP）机器学习被认为是在保护隐私的前提下利用敏感数据训练模型的黄金标准解决方案。然而，实现这一理想目标的主要障碍是其隐私-准确率权衡效果不佳，这一问题在DP表示学习中尤为突出。具体而言，已有研究表明，在中等隐私预算下，大多数模型学习到的表示并不比手工设计的特征显著更优。在本研究中，我们证明有效的DP表示学习可以通过图像字幕实现，并可扩展至互联网规模的多模态数据集。通过一系列工程技巧，我们使用合理的计算量，在LAION-2B的2.33亿子集上从头训练了一个DP图像字幕模型（DP-Cap），获得了前所未有的高质量图像特征，可用于多种下游视觉和视觉-语言任务。例如，在LAION数据集的隐私预算ε=8的条件下，基于学习到的DP-Cap特征训练的线性分类器在ImageNet-1K上达到65.8%的准确率，大幅提升了此前56.5%的最优水平（SOTA）。
英文摘要：
Differentially private (DP) machine learning is considered the gold-standard solution for training a model from sensitive data while still preserving privacy. However, a major barrier to achieving this ideal is its sub-optimal privacy-accuracy trade-off, which is particularly visible in DP representation learning. Specifically, it has been shown that under modest privacy budgets, most models learn representations that are not significantly better than hand-crafted features. In this work, we show that effective DP representation learning can be done via image captioning and scaling up to internet-scale multimodal datasets. Through a series of engineering tricks, we successfully train a DP image captioner (DP-Cap) on a 233M subset of LAION-2B from scratch using a reasonable amount of computation, and obtaining unprecedented high-quality image features that can be used in a variety of downstream vision and vision-language tasks. For example, under a privacy budget of $\varepsilon=8$ for the LAION dataset, a linear classifier trained on top of learned DP-Cap features attains $65.8\%$ accuracy on ImageNet-1K, considerably improving the previous SOTA of $56.5\%$.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
