INSPIRE: Incorporating Diverse Feature Preferences in Recourse
Authors: Prateek Yadav, Peter Hase, Mohit Bansal
TMLR 2024 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on popular real-world datasets demonstrates that our method is more fair compared to baselines and satisfies up to 25.9% more users. We also show that our method is robust to misspecifications of the cost function distribution.1 |
| Researcher Affiliation | Academia | Prateek Yadav EMAIL Peter Hase EMAIL Mohit Bansal EMAIL UNC-Chapel Hill |
| Pseudocode | Yes | Algorithm 1 Cost-Optimized Local Search Algorithm |
| Open Source Code | Yes | 1Our code is available at https://github.com/prateeky2806/EMC-COLS-recourse. |
| Open Datasets | Yes | To evaluate the effectiveness of EMC and COLS, we run experiments on two popular real-world datasets: Adult-Income (Dua & Graff, 2017) and COMPAS (Larson et al., 2016). |
| Dataset Splits | Yes | The data statistics for all the datasets can be found in Table 5. Table 5: Table containing data statistics and black-box model details. ... Train/val/test 20088/2338/749 1415/229/491 13172/1569/748 5491/705/444 |
| Hardware Specification | Yes | We ran our experiments on a local server using a single Nvidia 1080 Ti GPU. |
| Software Dependencies | No | The paper does not provide specific software names with version numbers. While it mentions Python libraries in the context of other methods (e.g., CARLA), it does not list its own dependencies with version numbers for reproducibility. |
| Experiment Setup | Yes | We set a fixed budget of 5000 model queries, a set size |S| = 10, and the number of cost function samples M = 1000 for all methods. ... Our black-box model is an Multi-Layer Perceptron with 2-layers. |