Sorting Out Lipschitz Function Approximation
Authors: Cem Anil, James Lucas, Roger Grosse
ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirically, we show that norm-constrained Group Sort networks achieve tighter estimates of Wasserstein distance than their Re LU counterparts and can achieve provable adversarial robustness guarantees with little cost to accuracy.Our experiments had two main goals. |
| Researcher Affiliation | Academia | 1Department of Computer Science, University of Toronto, Toronto, Canada 2Vector Institute, Toronto, Canada. |
| Pseudocode | No | The paper describes methods mathematically and in prose but does not include any formal pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not contain an explicit statement about open-sourcing the code for the described methodology or a direct link to a code repository. |
| Open Datasets | Yes | We present additional results in Appendix G, including CIFAR10 (Krizhevsky, 2009) classification and MNIST small data classification. We trained a GAN variant on MNIST and CIFAR10 datasets, then froze the weights of the generators. |
| Dataset Splits | No | The paper mentions training and test sets but does not provide specific details on the train/validation/test splits or how the data was partitioned. |
| Hardware Specification | Yes | We train our networks on a single NVIDIA 1080Ti GPU. |
| Software Dependencies | No | The paper does not provide specific version numbers for software dependencies used in its experiments. |
| Experiment Setup | Yes | Hyperparameters for training the critic are learning rate of 10^-4, batch size of 64, and Adam optimizer with β1 = 0.5, β2 = 0.9. We train for 100 epochs. |