In the social sciences, it is often necessary to debias studies and surveys before valid conclusions can be drawn. Debiasing algorithms enable the computational removal of bias using sample weights. However, an issue arises when only a subset of features is highly biased, while the rest is already representative. Algorithms need to strongly alter the sample distribution to manage a few highly biased features, which can in turn introduce bias into already representative variables. To address this issue, we developed a method that uses feature weights to minimize the impact of highly biased features on the computation of sample weights. Our algorithm is based on Maximum Representative Subsampling (MRS), which debiases datasets by aligning a non-representative sample with a representative one through iterative removal of elements to create a representative subsample. The new algorithm, named feature-weighted MRS (FW-MRS), decreases the emphasis on highly biased features, allowing it to retain more instances for downstream tasks. The feature weights are derived from the feature importance of a domain classifier trained to differentiate between the representative and non-representative datasets. We validated FW-MRS using eight tabular datasets, each of which we artificially biased. Biased features can be important for downstream tasks, and focusing less on them could lead to a decline in generalization. For this reason, we assessed the generalization performance of FW-MRS on downstream tasks and found no statistically significant differences. Additionally, FW-MRS was applied to a real-world dataset from the social sciences. The source code is available at https://github.com/kramerlab/FeatureWeightDebiasing.
Figures & tables
Figure 1: Schema for feature-weighted maximum representative sampling (FW-MRS). Two surveys stem from the same population: one is biased and includes the variable under investigation, while the other is representative but does not include it. FW-MRS mitigates bias by comparing the distributions of the biased and representative datasets using a classifier that leverages auxiliary information from the representative dataset to remove samples from the biased dataset. The algorithm returns a representative subset and feature weights that align the non-representative study to the distribution of the representative one.
Figure 2: Validation AUROC vs. number of dropped samples: All hyperparameters were fixed except for the temperature, which was varied. Each point denotes the mean AUROC across runs, with ellipses indicating the standard deviation. Circles correspond to FW-MRS RF and the triangle represents MRS.
Figure 3: Relative dropped samples and AUROC for 50 iterations with 10 times repeated 5-fold cross-validation for FW-MRS RF . Subfigure a) shows the relative dropped samples, and b) the corresponding AUROC. Hyperparameters were optimized on N for the downstream task, and the effect of different temperature settings is compared.
Figure 4: Relative dropped samples over 50 iterations with 10 times repeated 5-fold cross-validation.
Figure 5: Feature weights used in FW-MRS RF debiasing of GBS with auxiliary information of Allensbach.
Temperature
Number of Dropped Samples
MMD
MRS
501.1±4.9
0.153±0.004
0.5
501.0±3.6
0.151±0.004
0.25
497.5±3.6
0.145±0.003
0.1
493.2±4.2
0.132±0.005
0.05
478.7±5.1
0.114±0.004
0.025
444.5±4.9
0.086±0.002
Table 2: Mean number of dropped samples and MMD of 50 runs debiasing of GBS with auxiliary information of Allensbach with FW-MRS RF .
Appendix figures & tables2 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 6: Comparison of mean AUROC of MRS vs. FW-MRS RF with different temperatures on 50 debiasing runs of GBS with auxiliary information of Allensbach.
Figure 7: Feature importance used in FW-MRS RF debiasing of GBS with auxiliary information of Allensbach.
Real-world datasets often contain spurious correlations that are not causally related to the target label. When such correlations dominate the majority of training samples, models tend to rely on them, leading to misclassification of minority samples that do not exhibit the same spurious patterns. While a potential approach is to select subsets of data to better represent the minority samples, this may require access to group labels, which are typically unknown. Furthermore, as we demonstrate, widely used sample scoring functions in the invariant subset or coreset selection literature largely depend on spurious features and therefore fail to accurately capture the importance or difficulty of core, causally relevant features. Accordingly, we propose to mitigate spurious correlations by developing a two-stage sample scoring function that disentangles the learning dynamics of core and spurious features and evaluates their difficulty separately. Based on our proposed metric, we introduce a new algorithm to find and prioritize informative samples both with and without spurious correlations. Extensive experiments demonstrate that a standard ERM model trained on our selected samples achieves superior performance compared to state-of-the-art debiasing techniques, while requiring as little as 10% of the original training data.
Arda Fazla, Abolfazl Hashemi
School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 47907, USA.
Large-scale online marketplaces and recommender systems serve as critical technological support for e-commerce development. In industrial recommender systems, features play vital roles as they carry information for downstream models. Accurate feature importance estimation is critical because it helps identify the most useful feature subsets from thousands of feature candidates for online services. Such selection enables improved online performance while reducing computational cost. To address feature selection problems in deep learning, trainable gate-based and sensitivity-based methods have been proposed and proven effective in industrial practice. However, through the analysis of real-world cases, we identified three bias issues that cause feature importance estimation to rely on partial model layers, samples, or gradients, ultimately leading to inaccurate importance estimation. We refer to these as layer bias, baseline bias, and approximation bias. To mitigate these issues, we propose FairFS, a fair and accurate feature selection algorithm. FairFS regularizes feature importance estimated across all nonlinear transformation layers to address layer bias. It also introduces a smooth baseline feature close to the classifier decision boundary and adopts an aggregated approximation method to alleviate baseline and approximation biases. Extensive experiments demonstrate that FairFS effectively mitigates these biases and achieves state-of-the-art feature selection performance.
The issue of algorithmic biases in deep learning has led to the development of various debiasing techniques, many of which perform complex training procedures or dataset manipulation. However, an intriguing question arises: is it possible to extract fair and bias-agnostic subnetworks from standard vanilla-trained models without relying on additional data, such as unbiased training set? In this work, we introduce Bias-Invariant Subnetwork Extraction (BISE), a learning strategy that identifies and isolates "bias-free" subnetworks that already exist within conventionally trained models, without retraining or finetuning the original parameters. Our approach demonstrates that such subnetworks can be extracted via pruning and can operate without modification, effectively relying less on biased features and maintaining robust performance. Our findings contribute towards efficient bias mitigation through structural adaptation of pre-trained neural networks via parameter removal, as opposed to costly strategies that are either data-centric or involve (re)training all model parameters. Extensive experiments on common benchmarks show the advantages of our approach in terms of the performance and computational efficiency of the resulting debiased model.
Ivan Luiz De Moura Matos, Abdel Djalil Sad Saoud, Ekaterina Iakovleva +2
LTCI, Télécom Paris, Institut Polytechnique de Paris, France · MaLGa-DIBRIS, University of Genoa, Italy · AI for Good (AIGO), Istituto Italiano di Tecnologia, Italy