1,721,094 research outputs found
General Fair Empirical Risk Minimization
No full textWe tackle the problem of algorithmic fairness, where the goal is to avoid the unfairly influence of sensitive information, in the general context of regression with possible continuous sensitive attributes. We extend the framework of fair empirical risk minimization of [1] to this general scenario, covering in this way the whole standard supervised learning setting. Our generalized fairness measure reduces to well known notions of fairness available in literature. We derive learning guarantees for our method, that imply in particular its statistical consistency, both in terms of the risk and the fairness measure. We then specialize our approach to kernel methods and propose a convex fair estimator in that setting. We test the estimator on a commonly used benchmark dataset (Communities and Crime) and on a new dataset collected at the University of Genoa1, containing the information of the academic career of five thousand students. The latter dataset provides a challenging real case scenario of unfair behaviour of standard regression methods that benefits from our methodology. The experimental results show that our estimator is effective at mitigating the trade-off between accuracy and fairness requirements
Pac-Bayes and fairness: Risk and fairness bounds on distribution dependent fair priors
No full textWe address the problem of algorithmic fairness: ensuring that sensitive information does not unfairly influence the outcome of a classifier. We face this issue in the PAC-Bayes framework and we present an approach which trades off and bounds the risk and the fairness of the Gibbs Classifier measured with respect to different state-of-the-art fairness measures. For this purpose, we further develop the idea that the PAC-Bayes prior can be defined based on the data-generating distribution without actually needing to know it. In particular, we define a prior and a posterior which gives more weight to functions which exhibit good generalization and fairness properties
Convergence Properties of Stochastic Hypergradients
Full text linkBilevel optimization problems are receiving increasing attention in machine learning as they provide a natural framework for hyperparameter optimization and meta-learning. A key step to tackle these problems is the efficient computation of the gradient of the upper-level objective (hypergradient). In this work, we study stochastic approximation schemes for the hypergradient, which are important when the lower-level problem is empirical risk minimization on a large dataset. The method that we propose is a stochastic variant of the approximate implicit differentiation approach in (Pedregosa, 2016). We provide bounds for the mean square error of the hypergradient approximation, under the assumption that the lower-level problem is accessible only through a stochastic mapping which is a contraction in expectation. In particular, our main bound is agnostic to the choice of the two stochastic solvers employed by the procedure. We provide numerical experiments to support our theoretical analysis and to show the advantage of using stochastic hypergradients in practice
Randomized learning and generalization of fair and private classifiers: From PAC-Bayes to stability and differential privacy
No full textWe address the problem of randomized learning and generalization of fair and private classifiers. From one side we want to ensure that sensitive information does not unfairly influence the outcome of a classifier. From the other side we have to learn from data while preserving the privacy of individual observations. We initially face this issue in the PAC-Bayes framework presenting an approach which trades off and bounds the risk and the fairness of the randomized (Gibbs) classifier. Our new approach is able to handle several different state-of-the-art fairness measures. For this purpose, we further develop the idea that the PAC-Bayes prior can be defined based on the data-generating distribution without actually knowing it. In particular, we define a prior and a posterior which give more weight to functions with good generalization and fairness properties. Furthermore, we will show that this randomized classifier possesses interesting stability properties using the algorithmic distribution stability theory. Finally, we will show that the new posterior can be exploited to define a randomized accurate and fair algorithm. Differential privacy theory will allow us to derive that the latter algorithm has interesting privacy preserving properties ensuring our threefold goal of good generalization, fairness, and privacy of the final model
Multi-Rater Consensus Learning for Modeling Multiple Sparse Ratings of Affective Behaviour
Full text linkThe use of multiple raters to label datasets is an established practice in affective computing. The principal goal is to reduce unwanted subjective bias in the labelling process. Unfortunately, this leads to the key problem of identifying a ground truth for training the affect recognition system. This problem becomes more relevant in a sparsely-crossed annotation where each rater only labels a portion of the full dataset to ensure a manageable workload per rater. In this paper, we introduce a Multi-Rater Consensus Learning (MRCL) method which learns a representative affect recognition model that accounts for each rater's agreement with the other raters. MRCL combines a multitask learning (MTL) regularizer and a consensus loss. Unlike standard MTL, this approach allows the model to learn to predict each rater's label while explicitly accounting for the consensus among raters. We evaluated our approach on two different datasets based on spontaneous affective body movement expressions for pain behaviour detection and laughter type recognition respectively. The two naturalistic datasets were chosen for the different forms of labelling (different in affect, observation stimuli, and raters) that they together offer for evaluating our approach. Empirical results demonstrate that MRCL is effective for modelling affect from datasets with sparsely-crossed multi-rater annotation
Learning fair and transferable representations with theoretical guarantees
No full textDeveloping learning methods which do not discriminate subgroups in the population is the central goal of algorithmic fairness. One way to reach this goal is by modifying the data representation in order to satisfy prescribed fairness constraints. This allows to reuse the same representation in other context (tasks) without discriminate subgroups. In this work we measure fairness according to demographic parity, requiring the probability of the possible model decisions to be independent of the sensitive information. We argue that the goal of imposing demographic parity can be substantially facilitated within a multi-task learning setting. We leverage task similarities by encouraging a shared fair representation across the tasks via low rank matrix factorization. We derive learning bounds establishing that the learned representation transfers well to novel tasks both in terms of prediction performance and fairness metrics. We present experiments on three real world datasets, showing that the proposed method outperforms state-of-the-art approaches by a significant margin
Leveraging labeled and unlabeled data for consistent fair binary classification
No full textWe study the problem of fair binary classification using the notion of Equal Opportunity. It requires the true positive rate to distribute equally across the sensitive groups. Within this setting we show that the fair optimal classifier is obtained by recalibrating the Bayes classifier by a group-dependent threshold. We provide a constructive expression for the threshold. This result motivates us to devise a plug-in classification procedure based on both unlabeled and labeled datasets. While the latter is used to learn the output conditional probability, the former is used for calibration. The overall procedure can be computed in polynomial time and it is shown to be statistically consistent both in terms of the classification error and fairness measure. Finally, we present numerical experiments which indicate that our method is often superior or competitive with the state-of-the-art methods on benchmark datasets
- …
