Search

Supplement 1

Fair machine learning aims to mitigate the biases of model predictions against certain subpopulations regarding sensitive attributes such as race and gender. Among the many existing fairness notions, counterfactual fairness measures the model fairness from a causal perspective by comparing the predictions of each individual from the original data and the counterfactuals. In counterfactuals, the sensitive attribute values of this individual had been modified. Recently, a few works extend counterfactual fairness to graph data, but most of them neglect the following facts that can lead to biases: 1) the sensitive attributes of each node's neighbors may causally affect the prediction w.r.t. this node; 2) the sensitive attributes may causally affect other features and the graph structure. To tackle these issues, in this paper, we propose a novel fairness notion - graph counterfactual fairness, which considers the biases led by the above facts. To learn node representations towards graph counterfactual fairness, we propose a novel framework based on counterfactual data augmentation. In this framework, we generate counterfactuals corresponding to perturbations on each node's and their neighbors' sensitive attributes. Then we enforce fairness by minimizing the discrepancy between the representations learned from the original graph and the counterfactuals for each node. Experiments on both synthetic and real-world graphs show that our framework outperforms the state-of-the-art baselines in graph counterfactual fairness, and also achieves comparable prediction performance., Comment: 9 pages, 4 figures

To explore the complex underwater world, autonomous underwater vehicles (AUVs) are extensively used, whose recovery has been a challenging issue. To achieve high-efficiency AUV recovery, we propose a novel transformable underwater vehicle (TUV) which can maintain a closed state for low-resistance approaching and transform into an opening state for recovery to increase the possibility of success. The configuration of the thrusters is evaluated and optimized. The design and simulation results will be shown in this paper.

This work reports halogenated 5-(2-hydroxyphenyl)pyrazoles as pseudilin analogues with the potential to target the enzyme IspD in the methylerythritol phosphate (MEP) pathway. Such analogues were designed using the bioisosteric replacement of the pseudilin core structure and synthesized via an efficient three-step route. With AtIspD-based screening and pre- and post-emergence herbicidal tests, these compounds were demonstrated to have considerable activities against AtIspD, with IC50 up to 3.27 μM, and against model plants rape and barnyard grass, with moderate to excellent activities. At a rate of 150 g/ha in the greenhouse test, three compounds exhibited higher or comparable herbicidal activities than pseudilin. Molecular docking of representative compounds into the allosteric site of AtIspD revealed a binding mode similar to that of pseudilin. The established bioisosterism and synthesis method in this work may serve as an important tool for the development of new herbicides and antimicrobials targeting IspD in the MEP pathway.

Knowledge of the association information between the attributes in a data set provides insight into the underlying structure of the data and explains the relationships (independence, synergy, redundancy) between the attributes and class (if present). Complex models learnt computationally from the data are more interpretable to a human analyst when such interdependencies are known. In this paper, we focus on mining two types of association information among the attributes - correlation information and interaction information for both supervised (class attribute present) and unsupervised analysis (class attribute absent). Identifying the statistically significant attribute associations is a computationally challenging task - the number of possible associations increases exponentially and many associations contain redundant information when a number of correlated attributes are present. In this paper, we explore efficient data mining methods to discover non-redundant attribute sets that contain significant association information indicating the presence of informative patterns in the data., Comment: 16 pages, 7 figures

Background We have previously shown that childhood-onset rheumatic diseases show aberrant patterns of gene expression that reflect pathology-associated co-expression networks. In this study, we used novel computational approaches to examine how disease-associated networks are altered in one of the most common rheumatic diseases of childhood, juvenile idiopathic arthritis (JIA). Methods Using whole blood gene expression profiles derived from children in a pediatric rheumatology clinical trial, we used a network approach to understanding the impact of therapy and the underlying biology of response/non-response to therapy. Results We demonstrate that therapy for JIA is associated with extensive re-ordering of gene expression networks, even in children who respond inadequately to therapy. Furthermore, we observe distinct differences in the evolution of specific network properties when we compare children who have been treated successfully with those who have inadequate treatment response. Conclusions Despite the inherent noisiness of whole blood gene expression data, our findings demonstrate how therapeutic response might be mapped and understood in pathologically informative cells in a broad range of human inflammatory diseases.