Generalized Strategic Classification and the Case of Aligned Incentives
Authors: Sagi Levanon, Nir Rosenfeld
ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conclude with a set of experiments that empirically demonstrate the utility of our approach. |
| Researcher Affiliation | Academia | Faculty of Computer Science, Technion Israel Institute of Technology, Haifa, Israel. |
| Pseudocode | Yes | Algorithm 1 NL Hard formulation and Algorithm 2 NL/GP Soft formulation in Appendix D.1 and D.2. |
| Open Source Code | Yes | Code is publically available at https://github.com/Sagi Levanon1/GSC. |
| Open Datasets | Yes | We use the Coats Shopping dataset from Schnabel et al. (2016). and From each distribution we sampled 50 samples for the train set and 1250 samples for the test set. |
| Dataset Splits | No | Appendix B.1 states 'From each distribution we sampled 50 samples for the train set and 1250 samples for the test set.' and Appendix B.2 states 'From each distribution we sampled 25 samples for the train set and 1250 samples for the test set.' There is no explicit mention of a 'validation set' or corresponding split percentages/counts for validation data, although cross-validation is mentioned for hyperparameter tuning. |
| Hardware Specification | No | The paper does not explicitly state the specific hardware used for running its experiments (e.g., GPU models, CPU types, or memory). |
| Software Dependencies | No | For the optimization we used the standard Adam optimizer (Kingma & Ba, 2017) and a learning rate of 0.05. The paper does not mention specific software versions for libraries or frameworks used in the implementation. |
| Experiment Setup | Yes | We tuned this hyper-parameter using cross-validation of 3 splits for λ P t0.01, 0.1, 1u. For the optimization we used the standard Adam optimizer (Kingma & Ba, 2017) and a learning rate of 0.05. Each model trained for 200{200{50 epochs with a batch size of 5{16{24 for the generalization experiment, the varying preference noise experiment and the PPE experiment respectively. |