A Stability-based Validation Procedure for Differentially Private Machine Learning

By Kamalika Chaudhuri, Staal Amund Vinterbo. 2013

89 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

AdaCliP: Adaptive Clipping for Private SGD (2019, methodology): “In output perturbation techniques, the machine learned model is trained completely and then the final model is appropriately changed by using exponential mechanism [22, 11, 24, 16] or by adding Laplace or Gaussian noise to the final model [25, 10, 12].”
Differentially Private Learning with Adaptive Clipping (2019, neutral): “A method based on data splitting was provided in (Chaudhuri et al., 2011), whereas one based on satisfying certain stability conditions was introduced in (Chaudhuri and Vinterbo, 2013).”
Towards Practical Differentially Private Convex Optimization (2019, background): “Previous work [8], [14], [15] has shown this to be a challenging task.”
Private selection from private candidates (2018, background): “Chaudhuri and Vinterbo [8] studied this problem formally under strong assumptions on the algorithm.”
Three Tools for Practical Differential Privacy (2018, background): “In private training privacy spending accumulates for every trained model [3, 7].”
The Large Margin Mechanism for Differentially Private Maximization (2014, methodology): “This is a very general algorithmic problem that arises in many applications, include private PAC learning [25] (choosing the most accurate classifier), private decision tree induction [21] (choosing the most informative split), private frequent itemset mining [5] (choosing the most frequent itemset)”
Differentially Private Learning with Adaptive Clipping (2019, methodology): “A method based on data splitting was provided in [7], whereas one based on satisfying certain stability conditions was introduced in [6].”
Differentially private model selection with penalized and constrained likelihood (2016, methodology): “Furthermore, the noisy minimization method in Chaudhuri and Vinterbo (2013) relies on global sensitivity of the estimate and validation procedures, and hence requires bounded data, whereas our method uses a novel high probability upper bound of the local sensitivity, which requires a less stringent…”
Bolt-on Differential Privacy for Scalable Stochastic Gradient Descent-based Analytics (2016, methodology): “We use the tuning algorithm described in the original paper of Chaudhuri, Monteleoni and Sarwate [13], though the methodology and experiments in the following are readily extended to other private tuning algorithms [14].”
Large Language Models Can Be Strong Differentially Private Learners (2021, background): “Model selection from hyperparameter tuning on private training and validation data incurs extra leakage (Liu & Talwar, 2019; Chaudhuri & Vinterbo, 2013; Papernot & Steinke, 2021).”
Principled Evaluation of Differentially Private Algorithms using DPBench (2015, background): “Some work on differentially private machine learning pre-processes the input dataset using non-private data transformation algorithms [3, 9].”