Eliminating Latent Discrimination: Train Then Mask
Authors: Soheil Ghili, Ehsan Kazemi, Amin Karbasi3672-3680
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate the performance of our algorithm on several realworld datasets and show how fairness for these datasets can be improved with a very small loss in accuracy. In this section, we compare the performance of the train-then-mask algorithm to a number of baselines on real-world scenarios. |
| Researcher Affiliation | Academia | Soheil Ghili School of Management Yale University Ehsan Kazemi Yale Institute for Network Science Yale University Amin Karbasi Yale Institute for Network Science Yale University |
| Pseudocode | Yes | Algorithm 1 The Train-Then-Mask Algorithm 1: Compute the optimal classifier h (x0, x1, , xℓ) over all available features x0, x1, , and xℓ. 2: Keep the sensitive feature x0 fixed (e.g., define x0 = 0) for all data points. 3: Find the value of τ such that it maximizes the accuracy of h (0, x1, , xℓ) + τ over the validation set. |
| Open Source Code | No | The paper does not contain any statement about making its source code publicly available or provide a link to a code repository for the methodology described. |
| Open Datasets | Yes | We use the Adult Income and German Credit datasets from UCI Repository (Asuncion and Newman 2007; Blake and Merz 1998), and COMPAS Recidivism Risk dataset (Pro Publica 2018). |
| Dataset Splits | No | The paper mentions using a 'validation set' to find the optimal value of τ but does not provide specific details on the dataset split percentages or sample counts for the validation, training, or test sets. |
| Hardware Specification | No | The paper discusses the types of classifiers used (linear SVMs and neural networks) but does not provide specific details about the hardware specifications (e.g., GPU/CPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions using 'linear SVMs' and 'neural networks' but does not specify any software dependencies with version numbers (e.g., Python, specific libraries, or frameworks with versions) required to replicate the experiments. |
| Experiment Setup | No | The paper specifies the use of linear SVMs and neural networks with three hidden layers but does not provide specific hyperparameters such as learning rates, batch sizes, number of epochs, or optimizer settings for training these models. |