MetaLearning with Graph Neural Networks

Graph Neural Networks (GNNs), a generalization of deep neural networks on graph data have been widely used in various domains, ranging from drug discovery to recommender systems. However, GNNs on such applications are limited when there are few available samples. Meta-learning has been an important framework to address the lack of samples in machine learning, and in recent years, researchers have started to apply meta-learning to GNNs. In this work, we provide a comprehensive survey of different metalearning approaches involving GNNs on various graph problems showing the power of using these two approaches together. We categorize the literature based on proposed architectures, shared representations, and applications. Finally, we discuss several exciting future research directions and open problems.

An Interview with Dr. Shipeng Yu, Winner of ACM SIGKDD 2021 Service Award

Shipeng Yu, Ph.D. is the recipient of the 2021 ACM SIGKDD Service Award, which is the highest service award in the field of knowledge discovery and data mining. Conferred annually on one individual or group in recognition of outstanding professional services and contributions to the field of knowledge discovery and data mining, Dr. Yu was honored for his years of service and many accomplishments as general chair of KDD 2017 and currently as sponsorship director for SIGKDD. Dr. Yu is Director of AI Engineering, Head of the Growth AI team at LinkedIn, the world's largest professional network. He sat down with SIGKDD Explorations to discuss how he first got involved in the KDD conference in 2006, the benefits and drawbacks of virtual conferences, his work at LinkedIn, and KDD's place in the field of machine learning, data science and artificial intelligence.

Surf or sleep?

Poor sleep habits may cause serious problems of mind and body, and it is a commonly observed issue for college students due to study workload as well as peer and social influence. Understanding its impact and identifying students with poor sleep habits matters a lot in educational management. Most of the current researches is either based on self-reports and questionnaires, suffering from small sample size and social desirability bias, or the methods used are not suitable for the education system. In this paper, we develop a general data-driven method for identifying students' sleep patterns according to their Internet access pattern stored in the education management system and explore its influence from various aspects. First, we design a Possion-based probabilistic mixture model to cluster students according to the distribution of bedtime and identify students who are used to stay up late. Second, we profile students from five aspects (including eight dimensions) based on campusbehavior data and build Bayesian networks to explore the relationship between behavioral characteristics and sleeping habits. Finally, we test the predictability of sleeping habits. This paper not only contributes to the understanding of student sleep from a cognitive and behavioral perspective but also presents a new approach that provides an effective framework for various educational institutions to detect the sleeping patterns of students.

Ethical Adversaries

Machine learning is being integrated into a growing number of critical systems with far-reaching impacts on society. Unexpected behaviour and unfair decision processes are coming under increasing scrutiny due to this widespread use and its theoretical considerations. Individuals, as well as organisations, notice, test, and criticize unfair results to hold model designers and deployers accountable. We offer a framework that assists these groups in mitigating unfair representations stemming from the training datasets. Our framework relies on two inter-operating adversaries to improve fairness. First, a model is trained with the goal of preventing the guessing of protected attributes' values while limiting utility losses. This first step optimizes the model's parameters for fairness. Second, the framework leverages evasion attacks from adversarial machine learning to generate new examples that will be misclassified. These new examples are then used to retrain and improve the model in the first step. These two steps are iteratively applied until a significant improvement in fairness is obtained. We evaluated our framework on well-studied datasets in the fairness literature - including COMPAS - where it can surpass other approaches concerning demographic parity, equality of opportunity and also the model's utility. We investigated the trade-offs between these targets in terms of model hyperparameters and also illustrated our findings on the subtle difficulties when mitigating unfairness and highlight how our framework can assist model designers.

An Empirical Comparison of Bias Reduction Methods on Real-World Problems in High-Stakes Policy Settings

Applications of machine learning (ML) to high-stakes policy settings - such as education, criminal justice, healthcare, and social service delivery - have grown rapidly in recent years, sparking important conversations about how to ensure fair outcomes from these systems. The machine learning research community has responded to this challenge with a wide array of proposed fairness-enhancing strategies for ML models, but despite the large number of methods that have been developed, little empirical work exists evaluating these methods in real-world settings. Here, we seek to fill this research gap by investigating the performance of several methods that operate at different points in the ML pipeline across four real-world public policy and social good problems. Across these problems, we find a wide degree of variability and inconsistency in the ability of many of these methods to improve model fairness, but postprocessing by choosing group-specific score thresholds consistently removes disparities, with important implications for both the ML research community and practitioners deploying machine learning to inform consequential policy decisions.

Two Kinds of Discrimination in AI-Based Penal Decision-Making

The famous COMPAS case has demonstrated the difficulties in identifying and combatting bias and discrimination in AI-based penal decision-making. In this paper, I distinguish two kinds of discrimination that need to be addressed in this context. The first is related to the well-known problem of inevitable trade-offs between incompatible accounts of statistical fairness, while the second refers to the specific standards of discursive fairness that apply when basing human decisions on empirical evidence. I will sketch the essential requirements of non-discriminatory action within the penal sector for each dimension. Concerning the former, we must consider the relevant causes of perceived correlations between race and recidivism in order to assess the moral adequacy of alternative standards of statistical fairness, whereas regarding the latter, we must analyze the specific reasons owed in penal trials in order to establish what types of information must be provided when justifying court decisions through AI evidence. Both positions are defended against alternative views which try to circumvent discussions of statistical fairness or which tend to downplay the demands of discursive fairness, respectively.

Introduction to The Special Section on Bias and Fairness in AI

Fairness in Artificial Intelligence rightfully receives a lot of attention these days. Many life-impacting decisions are being partially automated, including health-care resource planning decisions, insurance and credit risk predictions, recidivism predictions, etc. Much of work appearing on this topic within the Data Mining, Machine Learning and Artificial Intelligence community is focused on technological aspects. Nevertheless, fairness is much wider than this as it lies at the intersection of philosophy, ethics, legislation, and practical perspectives. Therefore, to fill this gap and bring together scholars of these disciplines working on fairness, the first workshop on Bias and Fairness in AI was held online on September 18, 2020 at the ECML-PKDD 2020 conference. This special section includes six articles presenting different perspectives on bias and fairness from different angles.

Blind Spots in AI

Decisions support systems (DSS) are used more and more to offer right information at the right time. Serendipity has been pro- posed to ensure that the experience is broad and engaging. However, only designing for serendipity might not be enough to avoid historical discrimination affecting your DSS. For this reason we argue to include equity when designing for serendipity.

On the Applicability of Machine Learning Fairness Notions

Machine Learning (ML) based predictive systems are increasingly used to support decisions with a critical impact on individuals' lives such as college admission, job hiring, child custody, criminal risk assessment, etc. As a result, fairness emerged as an important requirement to guarantee that ML predictive systems do not discriminate against specific individuals or entire sub-populations, in particular, minorities. Given the inherent subjectivity of viewing the concept of fairness, several notions of fairness have been introduced in the literature. This paper is a survey of fairness notions that, unlike other surveys in the literature, addresses the question of "which notion of fairness is most suited to a given real-world scenario and why?". Our attempt to answer this question consists in (1) identifying the set of fairness-related characteristics of the real-world scenario at hand, (2) analyzing the behavior of each fairness notion, and then (3) fitting these two elements to recommend the most suitable fairness notion in every specific setup. The results are summarized in a decision diagram that can be used by practitioners and policy makers to navigate the relatively large catalogue of ML fairness notions.

Adversarial Attacks and Defenses

Despite the recent advances in a wide spectrum of applications, machine learning models, especially deep neural networks, have been shown to be vulnerable to adversarial attacks. Attackers add carefully-crafted perturbations to input, where the perturbations are almost imperceptible to humans, but can cause models to make wrong predictions. Techniques to protect models against adversarial input are called adversarial defense methods. Although many approaches have been proposed to study adversarial attacks and defenses in different scenarios, an intriguing and crucial challenge remains that how to really understand model vulnerability? Inspired by the saying that "if you know yourself and your enemy, you need not fear the battles", we may tackle the challenge above after interpreting machine learning models to open the black-boxes. The goal of model interpretation, or interpretable machine learning, is to extract human-understandable terms for the working mechanism of models. Recently, some approaches start incorporating interpretation into the exploration of adversarial attacks and defenses. Meanwhile, we also observe that many existing methods of adversarial attacks and defenses, although not explicitly claimed, can be understood from the perspective of interpretation. In this paper, we review recent work on adversarial attacks and defenses, particularly from the perspective of machine learning interpretation. We categorize interpretation into two types, feature-level interpretation, and model-level interpretation. For each type of interpretation, we elaborate on how it could be used for adversarial attacks and defenses. We then briefly illustrate additional correlations between interpretation and adversaries. Finally, we discuss the challenges and future directions for tackling adversary issues with interpretation.