Concise Explanations of Neural Networks using Adversarial Training
Authors: Prasad Chalasani, Jiefeng Chen, Amrita Roy Chowdhury, Xi Wu, Somesh Jha
ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our second contribution is an empirical verification of phenomenon (a), which we show, somewhat surprisingly, occurs not only in 1-layer networks, but also DNNs trained on standard image datasets...In Section 6 we empirically demonstrate that this sparsification effect of 1(")-adversarial training holds not only for 1-layer networks (e.g. logistic regression models), but also for Deep Convolutional Networks used for image classification...We ran experiments on five standard public benchmark datasets |
| Researcher Affiliation | Collaboration | Prasad Chalasani 1 Jiefeng Chen 2 Amrita Roy Chowdhury 2 Somesh Jha 1 2 Xi Wu 3 1Xai Pient 2University of Wisconsin (Madison) 3Google. |
| Pseudocode | No | The paper does not contain any sections labeled "Pseudocode" or "Algorithm", nor does it present any structured code-like blocks. |
| Open Source Code | Yes | The code for all experiments is at this repository: https://github.com/jfc43/advex. |
| Open Datasets | Yes | Convolutional Neural Networks on public benchmark image datasets MNIST (Le Cun & Cortes, 2010) and Fashion-MNIST (Xiao et al., 2017), and (b) logistic regression models on the Mushroom and Spambase tabular datasets from the UCI Data Repository (Dheeru & Karra Taniskidou, 2017). |
| Dataset Splits | No | The paper does not explicitly provide specific percentages or counts for training, validation, and test dataset splits, nor does it detail a cross-validation setup. It refers to "natural test data" but no specific breakdown. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running experiments, such as GPU models, CPU types, or memory specifications. It only mentions general computing environments without specific hardware details. |
| Software Dependencies | No | The paper does not specify the versions of any software dependencies, libraries, or frameworks used for the experiments (e.g., Python version, PyTorch/TensorFlow version). |
| Experiment Setup | No | The paper states: "Details of the datasets and training methodology are in Sec. J.1 of the Supplement." This implies setup details are available, but they are not provided in the main text itself. The main text mentions training logistic regression models and convolutional neural networks, and varying parameters like " " and "λ" but doesn't give specific hyperparameter values or comprehensive training configurations within the main body. |