If You Like Shapley Then You’ll Love the Core
Authors: Tom Yan, Ariel D. Procaccia5751-5759
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We also perform experiments that corroborate these theoretical results and shed light on settings where the least core may be preferable to the Shapley value. ... In our experiments, we verify these theoretical results and confirm that our algorithm can compute the least core easily and that the nucleolus is difficult to compute. ... Our experiments are conducted on feature valuation and data valuation tasks. |
| Researcher Affiliation | Academia | 1 Carnegie Mellon University 2 Harvard University |
| Pseudocode | No | The paper does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide explicit statements or links indicating the release of source code for the described methodology. |
| Open Datasets | Yes | We choose three smaller-scale UCI datasets (Dua and Graff 2017) ... For the natural dataset, we use the dog-vs-fish classification dataset as in the work of Koh and Liang (2017) and Ghorbani and Zou (2019). ... The specific dataset we use is the Enron Dataset, as in previous work (Ghorbani and Zou 2019; Koh and Liang 2017). |
| Dataset Splits | No | The paper mentions using a "validation set" for the Enron dataset, but it does not specify the explicit split percentages or absolute sample counts for training, validation, and test sets. For other datasets like UCI, it focuses on test accuracy without detailing the split methodology for training and validation. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used to run the experiments, such as GPU models, CPU types, or memory. |
| Software Dependencies | No | The paper mentions software components like "logistic regression," "neural networks," "Naive Bayes," and "Inception network" but does not specify their version numbers or any other software dependencies with version details. |
| Experiment Setup | Yes | We examine the performance of existing algorithms that one would use. To compare, we elect to fix the sample complexity (the number of v(S) queries) that the algorithms are permitted to use. ... We experiment with a budget of 5K, 10K, 25K, 50K for samples as in a low-resource setting. ... In total, 1000 data points are used for training a Naive Bayes model... We randomly flip the label for twenty percent of the data and allot a budget of 5000 samples for computing the solution concepts. |