Sample based Explanations via Generalized Representers
Authors: Che-Ping Tsai, Chih-Kuan Yeh, Pradeep Ravikumar
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Additionally, we conduct empirical comparisons of different generalized representers on two image and two text classification datasets. |
| Researcher Affiliation | Collaboration | Che-Ping Tsai Machine Learning Department Carnegie Mellon University chepingt@andrew.cmu.edu Chih-Kuan Yeh Google Deepmind jason6582@gmail.com Pradeep Ravikumar Machine Learning Department Carnegie Mellon University pradeepr@cs.cmu.edu |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks explicitly labeled as such. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described. |
| Open Datasets | Yes | For image classification, we follow Pruthi et al. [6] and use MNIST [59] and CIFAR-10 [60] datasets. For text classification, we follow Yeh et al. [53] and use Toxicity4 and AGnews 5 datasets, which contain toxicity comments and news of different categories respectively. 4https://www.kaggle.com/c/jigsaw-toxic-comment-classification-challenge 5http://groups.di.unipi.it/gulli/AG_corpus_of_news_articles.html |
| Dataset Splits | No | The paper states 'each class containing around 6,000 training samples' and refers to '10 randomly selected testing samples', but does not provide explicit details about training/validation/test splits or a dedicated validation set. |
| Hardware Specification | No | The paper mentions the complexity of the models (e.g., parameter count for CNNs) but does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details with version numbers (e.g., library or solver names with version numbers). |
| Experiment Setup | No | The paper mentions that 'The CNNs we use for the four datasets comprise 3 layers' and model parameter counts. It refers to 'stochastic gradient descent updates' with 'mini-batch and the learning rate' and mentions Appendix C for more details, but does not provide specific hyperparameter values or detailed system-level training settings in the main text. |