In this paper, we study differentially private empirical risk minimization (DP-ERM). It has been shown that the worst-case utility of DP-ERM reduces polynomially as the dimension increases. This is a major obstacle to privately learning large machine learning models. In high dimension, it is common for some model's parameters to carry more information than others. To exploit this, we propose a differentially private greedy coordinate descent (DP-GCD) algorithm. At each iteration, DP-GCD privately performs a coordinate-wise gradient step along the gradients' (approximately) greatest entry. We show theoretically that DP-GCD can achieve a logarithmic dependence on the dimension for a wide range of problems by naturally exploiting their structural properties (such as quasi-sparse solutions). We illustrate this behavior numerically, both on synthetic and real datasets.
翻译:在本文中,我们研究了不同程度的私人经验风险最小化(DP-EMM),已经表明DP-EMM最差的效用随着尺寸的增加而减少了多元性。这是私人学习大型机器学习模型的主要障碍。在高维方面,某些模型参数通常会携带比其他参数更多的信息。为了利用这一点,我们建议采用差异化的私人贪婪协调血统(DP-GCD)算法。在每一次迭代中,DP-GCD私下沿着(约)最大的梯度条目执行一个协调的梯度步骤。我们从理论上表明,DP-GCD可以通过自然利用其结构特性(如准扭曲的解决方案)实现对广泛问题层面的逻辑依赖。我们在合成和真实数据集中都用数字来说明这一行为。