Fairness risk measures
Authors: Robert Williamson, Aditya Menon
ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate that the CVa R-fairness minimiser (27) empirically yields reasonable fairness-accuracy tradeoffs. We present results on a synthetic two-dimensional dataset (synth) from Donini et al. (2018), where there is a single binary sensitive feature S, and the UCI adult dataset with gender as the binary S. |
| Researcher Affiliation | Collaboration | 1Australian National University 2Google Research. |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks (clearly labeled algorithm sections or code-like formatted procedures). |
| Open Source Code | No | The paper does not provide concrete access to source code (specific repository link, explicit code release statement, or code in supplementary materials) for the methodology described in this paper. |
| Open Datasets | Yes | We present results on a synthetic two-dimensional dataset (synth) from Donini et al. (2018), where there is a single binary sensitive feature S, and the UCI adult dataset with gender as the binary S. |
| Dataset Splits | Yes | We use the validation procedure of Donini et al. (2018) to tune the regularisation strength, using balanced error as the base measure. This is repeated over 100 random 80/20% train-test splits. |
| Hardware Specification | No | The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers like Python 3.8, CPLEX 12.4) needed to replicate the experiment. |
| Experiment Setup | Yes | We use square-hinge ℓsh(y, f) = [1 − yf]2 + as our base loss, and regularised linear scorers as our F. We use the validation procedure of Donini et al. (2018) to tune the regularisation strength, using balanced error as the base measure. |