Learning Classifiers That Induce Markets
Authors: Yonatan Sommer, Ivri Hikri, Lotan Amit, Nir Rosenfeld
ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We now turn to demonstrate how our market-aware strategic learning framework performs empirically on real data with simulated market behavior. We use two datasets common and publicly available datasets and adapt them to our strategic market setting: (i) the adult dataset, showed here, and using capital_gain feature as a proxy for budgets b; and (ii) the folktables dataset, deferred to Appendix B.5. |
| Researcher Affiliation | Academia | 1Faculty of Computer Science, Technion Israel Institute of Technology, Haifa, Israel. Correspondence to: Nir Rosenfeld <EMAIL>. |
| Pseudocode | Yes | Algorithm 1 Exact empirical market prices |
| Open Source Code | Yes | Code is publicly available at https://github.com/BML-Technion/MASC. |
| Open Datasets | Yes | We use two datasets common and publicly available datasets and adapt them to our strategic market setting: (i) the adult dataset, showed here, and using capital_gain feature as a proxy for budgets b; and (ii) the folktables dataset, deferred to Appendix B.5. ... The data is publicly available at https: //archive.ics.uci.edu/dataset/2/adult. ... The data is publicly available at https://github.com/socialfoundations/folktables. |
| Dataset Splits | Yes | Data splits. We used a train-validation-test split of 70:10:20 and averaged the results over 10 random data splits. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, processor types, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | All code was implemented in python, and the learning framework was implemented using Pytorch. The paper does not provide specific version numbers for these software components. |
| Experiment Setup | Yes | We used the following hyperparameters: Temperature Tsoftsort for the softsort operator: 0.001 Temperature Tsoftmax for the softmax operator: 0.01 Batch size: 500 Learning rate: adult: 0.001 for budget_scale [1, 32], 0.01 for budget_scale [64, 1024] folktables: 0.001 for all budget scales Regularization and coefficient: 0.1 Epochs: 100 for adult, 1000 for folktables |