Making AI Forget You: Data Deletion in Machine Learning
Authors: Antonio Ginart, Melody Guan, Gregory Valiant, James Y. Zou
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5 Experiments, Datasets We run our experiments on five real, publicly available datasets: Celltype (N =12,009, D = 10, K = 4) [42], Covtype (N = 15,120, D = 52, K = 7) [12], MNIST (N = 60,000, D = 784, K =10) [51], Postures (N =74,975, D =15, K =5) [35, 34] , Botnet (N =1,018,298, D =115, K =11)[56], and a synthetic dataset made from a Gaussian mixture model which we call Gaussian (N =100,000, D=25, K =5). In Table 4, we report the amortized total runtime of training and deletion for each method. Overall, we see that the statistical clustering performance of the three methods are competitive. Furthermore, we find that both proposed algorithms yield orders of magnitude of speedup. |
| Researcher Affiliation | Academia | 1Dept. of Electrical Engineering 2Dept. of Computer Science 3Dept. of Biomedial Data Science Stanford University, Palo Alto, CA 94305 {tginart, mguan, valiant, jamesz}@stanford.edu |
| Pseudocode | Yes | Algorithm 1 Quantized k-means (abridged), Algorithm 2 DC-k-means |
| Open Source Code | No | The paper does not contain an unambiguous statement of releasing the source code for the work described in this paper, nor does it provide a direct link to a code repository. |
| Open Datasets | Yes | Datasets We run our experiments on five real, publicly available datasets: Celltype (N =12,009, D = 10, K = 4) [42], Covtype (N = 15,120, D = 52, K = 7) [12], MNIST (N = 60,000, D = 784, K =10) [51], Postures (N =74,975, D =15, K =5) [35, 34] , Botnet (N =1,018,298, D =115, K =11)[56], and a synthetic dataset made from a Gaussian mixture model which we call Gaussian (N =100,000, D=25, K =5). |
| Dataset Splits | No | The paper mentions training on the 'full dataset' and simulating deletion requests but does not provide specific training/validation/test dataset splits (percentages, absolute counts, or predefined splits) for model development or hyperparameter tuning. |
| Hardware Specification | No | The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper mentions using scikit-learn and numpy but does not provide specific version numbers for these or any other ancillary software dependencies needed to replicate the experiments. |
| Experiment Setup | Yes | For both of our proposed algorithms, we always fix 10 iterations of Lloyd s, and all other parameters are selected with simple but effective heuristics (see Appendix D). |