ACCURACY FIRST: SELECTING A DIFFERENTIAL PRIVACY LEVEL FOR ACCURACY-CONSTRAINED ERM

Keyphrases

• Empirical Risk Minimization 100%
• Differential Privacy 100%
• Privacy Level 100%
• Accuracy Constraints 50%
• Empirical Risk Minimisation Algorithm 50%
• Privacy Analysis 50%
• Accuracy Requirements 50%
• Noise Reduction 25%
• General Noise 25%
• Regularized Logistic Regression 25%
• Privacy Parameter 25%
• Given Accuracy 25%
• Fixed-accuracy 25%
• Utility Guarantees 25%
• Noise Reduction Algorithm 25%
• Regularized Linear Regression 25%
• Common Objective 25%
• Binary Search 25%
• Practical Setting 25%
• Privacy Requirements 25%
• Privacy Tools 25%
• Privacy Constraint 25%
• Data Analysts 25%

Computer Science

• Differential Privacy 100%
• Risk Minimization 100%
• Accuracy Requirement 50%
• Objective Function 25%
• Data Analyst 25%
• Privacy Requirement 25%
• Logistic Regression 25%