The shift between the training and testing distributions is commonly due to sample selection bias, a type of bias caused by non-random sampling of examples to be included in the training set. Although there are many approaches proposed to learn a classifier under sample selection bias, few address the case where a subset of labels in the training set are missing-not-at-random (MNAR) as a result of the selection process. In statistics, Greene’s method formulates this type of sample selection with logistic regression as the prediction model. However, we find that simply integrating this method into a robust classification framework is not effective for this bias setting. In this paper, we propose BiasCorr, an algorithm that improves on Greene’s method by modifying the original training set in order for a classifier to learn under MNAR sample selection bias. We provide theoretical guarantee for the improvement of BiasCorr over Greene’s method by analyzing its bias. Experimental results on real-world datasets demonstrate that BiasCorr produces robust classifiers and can be extended to outperform state-of-the-art classifiers that have been proposed to train under sample selection bias.

Personalized federated learning (PFL) gives each client in a federation the power to obtain a model tailored to their specific data distribution or task without the client forfeiting the benefits of training in a federated manner. However, the concept of demographic group fairness has not been widely studied in PFL. Further, fairness heterogeneity – when not all clients enforce the same local fairness metric – has not been studied at all. To fill this gap, we propose Fair Hypernetworks (FHN), a personalized federated learning architecture based on hypernetworks that is robust to statistical (e.g., non-IID and unbalanced data) and fairness heterogeneity. We theoretically show that granting clients the ability to independently choose multiple (possibly conflicting) fairness constraints, such as demographic parity or equalized odds, does not break previously proven generalization bounds on hypernetworks used in the federated setting. Additionally, we empirically test FHN against several baselines in multiple fair federated learning settings, and we find t hat F HN outperforms all other federated baselines when handling clients with heterogeneous fairness metrics. We further demonstrate the scalability of FHN to show that minimal degradation to the accuracy and the fairness of the clients occurs when the federation grows in size. Additionally, we empirically validate our theoretical analysis to show FHN generalizes well to new clients. To our knowledge, our FHN architecture is the first to consider tolerance to fairness heterogeneity which gives clients the freedom to personalize the fairness metric enforced during local training.

Recent research on fair regression focused on developing new fairness notions and approximation methods as target variables and even the sensitive attribute are continuous in the regression setting. However, all previous fair regression research assumed the training data and testing data are drawn from the same distributions. This assumption is often violated in real world due to the sample selection bias between the training and testing data. In this paper, we develop a framework for fair regression under sample selection bias when dependent variable values of a set of samples from the training data are missing as a result of another hidden process. Our framework adopts the classic Heckman model for bias correction and the Lagrange duality to achieve fairness in regression based on a variety of fairness notions. Heckman model describes the sample selection process and uses a derived variable called the Inverse Mills Ratio (IMR) to correct sample selection bias. We use fairness inequality and equality constraints to describe a variety of fairness notions and apply the Lagrange duality theory to transform the primal problem into the dual convex optimization. For the two popular fairness notions, mean difference and mean squared error difference, we derive explicit formulas without iterative optimization, and for Pearson correlation, we derive its conditions of achieving strong duality. We conduct experiments on three real-world datasets and the experimental results demonstrate the approach’s effectiveness in terms of both utility and fairness metrics.

Adversarial machine learning has been extensively studied from perspectives of attack settings and defense strategies. However, existing adversarial training models fail to be adaptive and robust against new attacks during test time. In this paper, we propose a novel adversarially adaptive defense (AAD) framework based on adaptive training such that the trained prediction and detection models adapt at test time to new attacks. Our AAD structures the training data into groups and each group represents one attack scenario. Different from empirical risk minimization that trains a single robust model or learns an invariant feature space, our AAD learns a context vector from features of each batch during training and incorporates the learned context vector into both prediction and detection models. Thus, AAD can adapt at test time to new adversarial attacks. We formulate our problem by optimizing a joint loss from prediction, detection, and regularization via a multi-task learning framework. We conduct comprehensive empirical evaluations with popular adversarial attacks and defense strategies on two real-world datasets under different attack settings. Empirical results show that AAD achieves both high prediction and detection accuracy and significantly outperforms baselines.

To address the sample selection bias between the training and test data, previous research works focus on reweighing biased training data to match the test data and then building classification models on the reweighed training data. However, how to achieve fairness in the built classification models is under-explored. In this paper, we propose a framework for robust and fair learning under sample selection bias. Our framework adopts the reweighing estimation approach for bias correction and the minimax robust estimation approach for achieving robustness on prediction accuracy. Moreover, during the minimax optimization, the fairness is achieved under the worst case, which guarantees the model's fairness on test data. We further develop two algorithms to handle sample selection bias when test data is both available and unavailable.

In differentially private stochastic gradient descent (DPSGD), gradient clipping and random noise addition disproportionately affect underrepresented and complex classes and subgroups. As a consequence, DPSGD has disparate impact: the accuracy of a model trained using DPSGD tends to decrease more on these classes and subgroups vs. the original, non-private model. If the original model is unfair in the sense that its accuracy is not the same across all subgroups, DPSGD exacerbates this unfairness. In this work, we study the inequality in utility loss due to differential privacy, which compares the changes in prediction accuracy w.r.t. each group between the private model and the non-private model. We analyze the cost of privacy w.r.t. each group and explain how the group sample size along with other factors is related to the privacy impact on group accuracy. Furthermore, we propose a modified DPSGD algorithm, called DPSGD-F, to achieve differential privacy, equal costs of differential privacy, and good utility. DPSGD-F adaptively adjusts the contribution of samples in a group depending on the group clipping bias such that differential privacy has no disparate impact on group accuracy. Our experimental evaluation shows the effectiveness of our removal algorithm on achieving equal costs of differential privacy with satisfactory utility.

Federated learning is an emerging framework that builds centralized machine learning models with training data distributed across multiple devices.Most of the previous works about federated learning focus on the privacy protection and communication cost reduction.However, how to achieve fairness in federated learning is underexplored and challenging especially when testing data distribution is different from training distribution or even unknown.Introducing simple fairness constraints on the centralized model cannot achieve model fairness on unknown testing data.In this paper, we develop a fairness-aware agnostic federated learning framework (Agnostic-Fair) to deal with the challenge of unknown testing distribution.We use kernel reweighing functions to assign a reweighing value on each training sample in both loss function and fairness constraint.Therefore, the centralized model built from AgnosticFair can achieve high accuracy and fairness guarantee on unknown testing data.Moreover, the built model can be directly applied to local sites as it guarantees fairness on local data distributions.To our best knowledge, this is the first work to achieve fairness in federated learning.Experimental results on two real datasets demonstrate the effectiveness in terms of both utility and fairness under data shift scenarios.

The data generation sources are increasing in the past few years, such as mobile devices, embedded sensors, various intelligent equipment and so forth. These increasing data sources push the deployment of deep learning models in a distributed manner. However, the traditional distributed deep learning is to build a global model over all collected data and may overlook specific components which are of vital importance to personalized users. In this paper, we propose a learning framework that allows an individual user to build a personalized model. Our framework consists of two stages, including efficient similar data selection from other users and adversarial training. Instead of selecting similar data by computing hand-designed similarity metrics, we train an auto-encoder and a GAN on individual user's data, and use them to request similar data from other users. To further improve the personalized model performance, we apply adversarial training to minimize the distribution discrepancy between requested data and user's own data. Experimental results demonstrate the effectiveness of the proposed framework.