Research map / Federated learning and privacy / Differential privacy
Differentially Private Regression and Classification with Sparse Gaussian Processes
Michael Thomas Smith, Mauricio A. Alvarez, Neil D. Lawrence
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add sparse inducing-input approximation to DP cloaking | DP GP regression · !Kung 1d (DP eps=1) | RMSE (cm): 13.3 -> 9.9 | Improved |
| add sparse inducing-input approximation to DP cloaking | DP GP regression · !Kung 2d (DP eps=1) | RMSE (cm): 17.2 -> 10.2 | Improved |
| add variable lengthscale to DP cloaking | DP GP regression · !Kung 1d (DP eps=1) | RMSE (cm): 13.3 -> 12.8 | No clear effect |
| add variable lengthscale to DP cloaking | DP GP regression · !Kung 2d (DP eps=1) | RMSE (cm): 17.2 -> 16.3 | No clear effect |
| add weight as a second input to standard DP cloaking | DP GP regression · !Kung (DP eps=1) | RMSE (cm): 13.3 -> 17.2 | Hurt |
| add sparse approximation to GP without DP | GP regression · !Kung 1d (no DP) | RMSE (cm): 7.1 -> 7.7 | Hurt |
| add sparse approximation to GP without DP | GP regression · !Kung 2d (no DP) | RMSE (cm): 5.2 -> 7.7 | Hurt |
| add a second Laplace optimisation iteration with split privacy budget (ablation) | DP GP classification · 2D stripe classification | accuracy (%): 69% to 51% | Hurt |
| use fewer inducing inputs and longer lengthscale (ablation) | DP sparse GP classifier · MNIST low/high (15x15, 256 train) | accuracy | Improved |
| select hyperparameters with exponential mechanism instead of random choice | DP GP regression hyperparameters · !Kung 1d (80 configurations) | expected RMSE (cm): 153.81 -> 19.02 | Improved |
| increase privacy (smaller epsilon) when selecting lengthscale (ablation) | DP GP hyperparameter selection · !Kung 1d | selected lengthscale | Mixed |
Closest papers by meaning
- The importance of feature preprocessing for differentially private linear optimization (ICLR 2024 · similarity 0.89)
- Instance-Specific Asymmetric Sensitivity in Differential Privacy (NeurIPS 2024 · similarity 0.89)
- Differentially Private High-dimensional Variable Selection via Integer Programming (NeurIPS 2025 · similarity 0.88)
- On the Computational Complexity of Private High-dimensional Model Selection (NeurIPS 2024 · similarity 0.88)
- Certification for Differentially Private Prediction in Gradient-Based Training (ICML 2025 · similarity 0.88)
- Shifted Interpolation for Differential Privacy (ICML 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
