Currently, various uncertainty quantification methods have been proposed to provide certainty and probability estimates for deep learning models' label predictions. Meanwhile, with the growing demand for the right to be forgotten, machine unlearning has been extensively studied as a means to remove the impact of requested sensitive data from a pre-trained model without retraining the model from scratch. However, the vulnerabilities of such generated predictive uncertainties with regard to dedicated malicious unlearning attacks remain unexplored. To bridge this gap, for the first time, we propose a new class of malicious unlearning attacks against predictive uncertainties, where the adversary aims to cause the desired manipulations of specific predictive uncertainty results. We also design novel optimization frameworks for our attacks and conduct extensive experiments, including black-box scenarios. Notably, our extensive experiments show that our attacks are more effective in manipulating predictive uncertainties than traditional attacks that focus on label misclassifications, and existing defenses against conventional attacks are ineffective against our attacks.

Currently, attention mechanism has become a standard fixture in most state-of-the-art natural language processing (NLP) models, not only due to the outstanding performance it could gain but also due to plausible innate explanations for the behaviors of neural architectures it provides, which is notoriously difficult to analyze. However, recent studies show that attention is unstable against randomness and perturbations during training or testing, such as random seeds and slight perturbation of embedding vectors, which impedes it from becoming a faithful explanation tool. Thus, a natural question is whether we can find some substitute for the current attention that is more stable and could keep the most important characteristics of explanation and prediction of attention. In this paper, to resolve the problem, we provide a rigorous definition of such alternate namely SEAT (Stable andExplainableAttention). Specifically, a SEAT should have the following three properties: (1) Its prediction distribution is enforced to be close to the distribution based on the vanilla attention; (2) Its top-$k$indices have large overlaps with those of the vanilla attention; (3) It is robust w.r.t perturbations, i.e., any slight perturbation on SEAT will not change the prediction distribution too much, which implicitly indicates that it is stable to randomness and perturbations. To further improve the interpretability stability against perturbations, based on SEAT we provide another definition called SEAT++. Then we propose a method to get a SEAT++, which could be considered an ad hoc modification for canonical attention. Finally, through intensive experiments on various datasets, we compare our SEAT and SEAT++ with other baseline methods using RNN, BiLSTM, and BERT architectures via six different evaluation metrics for model interpretation, stability, and accuracy. Results show that SEAT and SEAT++ are more stable against different perturbations and randomness while also keeping the explainability of attention, which indicates they provide more faithful explanations. Moreover, compared with vanilla attention, there is almost no utility (accuracy) degradation for SEAT and SEAT++.

Synthesizing electronic health records (EHR) data has become a preferred strategy to address data scarcity, improve data quality, and model fairness in healthcare. However, existing approaches for EHR data generation predominantly rely on state-of-the-art generative techniques like generative adversarial networks, variational autoencoders, and language models. These methods typically replicate input visits, resulting in inadequate modeling of temporal dependencies between visits and overlooking the generation of time information, a crucial element in EHR data. Moreover, their ability to learn visit representations is limited due to simple linear mapping functions, thus compromising generation quality. To address these limitations, we propose a novel EHR data generation model called EHRPD. It is a diffusion-based model designed to predict the next visit based on the current one while also incorporating time interval estimation. To enhance generation quality and diversity, we introduce a novel time-aware visit embedding module and a pioneering predictive denoising diffusion probabilistic model (P-DDPM). Additionally, we devise a predictive U-Net (PU-Net) to optimize P-DDPM. We conduct experiments on two public datasets and evaluate EHRPD from fidelity, privacy, and utility perspectives. The experimental results demonstrate the efficacy and utility of the proposed EHRPD in addressing the aforementioned limitations and advancing EHR data generation.

Health risk prediction aims to forecast the potential health risks that patients may face using their historical Electronic Health Records (EHR). Although several effective models have developed, data insufficiency is a key issue undermining their effectiveness. Various data generation and augmentation methods have been introduced to mitigate this issue by expanding the size of the training data set through learning underlying data distributions. However, the performance of these methods is often limited due to their task-unrelated design. To address these shortcomings, this paper introduces a novel, end-to-end diffusion-based risk prediction model, named MedDiffusion. It enhances risk prediction performance by creating synthetic patient data during training to enlarge sample space. Furthermore, MedDiffusion discerns hidden relationships between patient visits using a step-wise attention mechanism, enabling the model to automatically retain the most vital information for generating high-quality data. Experimental evaluation on four real-world medical datasets demonstrates that MedDiffusion outperforms 14 cutting-edge baselines in terms of PR-AUC, F1, and Cohen's Kappa. We also conduct ablation studies and benchmark our model against GAN-based alternatives to further validate the rationality and adaptability of our model design. Additionally, we analyze generated data to offer fresh insights into the model's interpretability. The source code is available via https://shorturl.at/aerT0.

Meta-learning has shown great promise in few-shot image classification where only a small amount of labeled data is available in each classification task. Many training tasks are provided to train a meta-model that can quickly learn new and similar concepts with few labeled samples. Data augmentation is often used to augment training tasks to avoid overfitting. However, existing data augmentation methods are often manually designed and fixed during training, ignoring training dynamics and the difference between various meta-learning settings specified by meta-model architectures and meta-learning algorithms. To address this problem, we add a task transformation layer between a training task and a meta-model such that the right amount of perturbation is added to training tasks for a certain meta-learning setting at a certain training stage. By jointly optimizing the task transformation layer and the meta-model, we avoid the risk of providing tasks that are either too easy or too difficult during training. We design the task transformation layer as a stochastic transformation function, adding the flexibility in how a training task can be transformed. We leverage differentiable data augmentations as the building blocks of the task transformation function for efficient optimization. Extensive experiments show that our method can consistently improve the few-shot generalization performance of various meta-models trained with different meta-learning algorithms, meta-model architectures, and datasets.

With ubiquitous adoption of machine learning algorithms in web technologies, such as recommendation system and social network, algorithm fairness has become a trending topic, and it has a great impact on social welfare. Among different fairness definitions, path-specific causal fairness is a widely adopted one with great potentials, as it distinguishes the fair and unfair effects that the sensitive attributes exert on algorithm predictions. Existing methods based on path-specific causal fairness either require graph structure as the prior knowledge or have high complexity in the calculation of path-specific effect. To tackle these challenges, we propose a novel casual graph based fair prediction framework which integrates graph structure learning into fair prediction to ensure that unfair pathways are excluded in the causal graph. Furthermore, we generalize the proposed framework to the scenarios where sensitive attributes can be non-root nodes and affected by other variables, which is commonly observed in real-world applications, such as recommendation system, but hardly addressed by existing works. We provide theoretical analysis on the generalization bound for the proposed fair prediction method, and conduct a series of experiments on real-world datasets to demonstrate that the proposed framework can provide better prediction performance and algorithm fairness trade-off.

With the increasing growth of data and the ability of learning with them, machine learning models are adopted in various domains. However, few of machine learning models are able to reason their prediction, which limits their further applications in real-world tasks. With the potential to address this dilemma, model interpretation has become an important research topic because of the ability to provide the underlying reasons for model predictions at the feature level or concept level. Model interpretation at the concept level focuses on exploring the roles of concepts in model prediction, which enables more compact and understandable interpretations. Concept-level model interpretation requires the identification of the concepts that contribute to model prediction and the exploration of the rules underneath these concepts. To achieve the two objectives, we propose a Concept-level Model Interpretation framework (CMIC) from the perspective of causality. CMIC can automatically detect concepts in data and discover the causal relation between the detected concepts and the model's predicted labels. Furthermore, CMIC ranks the contributions of concepts by their causal effect on the model prediction, reflecting the detected concepts’ importance. We evaluate the proposed CMIC framework on both synthetic and real-world datasets to demonstrate the quality of the provided interpretation.

This paper studies the convergence and generalization of a large class of Stochastic Gradient Descent (SGD) momentum schemes, in both learning from scratch and transferring representations with fine-tuning. Momentum-based acceleration of SGD is the default optimizer for many deep learning models. However, there is a lack of general convergence guarantees for many existing momentum variants in conjunction withstochastic gradient. It is also unclear how the momentum methods may affect thegeneralization error. In this paper, we give a unified analysis of several popular optimizers, e.g., Polyak's heavy ball momentum and Nesterov's accelerated gradient. Our contribution is threefold. First, we give a unified convergence guarantee for a large class of momentum variants in thestochastic setting. Notably, our results cover both convex and nonconvex objectives. Second, we prove a generalization bound for neural networks trained by momentum variants. We analyze how hyperparameters affect the generalization bound and consequently propose guidelines on how to tune these hyperparameters in various momentum schemes to generalize well. We provide extensive empirical evidence to our proposed guidelines. Third, this study fills the vacancy of a formal analysis of fine-tuning in literature. To our best knowledge, our work is the first systematic generalizability analysis on momentum methods that cover both learning from scratch and fine-tuning. Our codes are available https://github.com/jsycsjh/Demystify-Hyperparameters-for-Stochastic-Optimization-with-Transferable-Representations .

Metric learning aims at automatically learning a distance metric from data so that the precise similarity between data instances can be faithfully reflected, and its importance has long been recognized in many fields. An implicit assumption in existing metric learning works is that the learned models are performed in a reliable and secure environment. However, the increasingly critical role of metric learning makes it susceptible to a risk of being malicious attacked. To well understand the performance of metric learning models in adversarial environments, in this article, we study the robustness of metric learning to adversarial perturbations, which are also known as the imperceptible changes to the input data that are crafted by an attacker to fool a well-learned model. However, different from traditional classification models, metric learning models take instance pairs rather than individual instances as input, and the perturbation on one instance may not necessarily affect the prediction result for an instance pair, which makes it more difficult to study the robustness of metric learning. To address this challenge, in this article, we first provide a definition of pairwise robustness for metric learning, and then propose a novel projected gradient descent-based attack method (called AckMetric) to evaluate the robustness of metric learning models. To further explore the capability of the attacker to change the prediction results, we also propose a theoretical framework to derive the upper bound of the pairwise adversarial loss. Finally, we incorporate the derived bound into the training process of metric learning and design a novel defense method to make the learned models more robust. Extensive experiments on real-world datasets demonstrate the effectiveness of the proposed methods.

Inferring causal effect from observational data has attracted much attention from various domains. Under the potential outcome framework, the estimation of counterfactuals is crucial for the investigation of causal effect at the individual level. Existing representation learning approaches focus on learning one balanced feature space, which ignores certain information predictive to the outcomes. To fully utilize the predictive information, we propose a Subspace learning based Counterfactual Inference (SCI) method to estimate causal effect at the individual level. Different from existing work, SCI learns both a common subspace, which preserves the information across all the treatment groups, and treatment-specific subspaces, which retain the information associated with each specific treatment. Learning from two kinds of subspaces helps SCI obtain better causal effect estimations than state-of-the-art methods, demonstrated by a series of experiments on synthetic and real-world datasets.

The past years have witnessed the rapid development of deep reinforcement learning (DRL), which is a combination of deep learning and reinforcement learning (RL). However, the adoption of deep neural networks makes the decision-making process of DRL opaque and lacking transparency. Motivated by this, various interpretation methods for DRL have been proposed. However, those interpretation methods make an implicit assumption that they are performed in a reliable and secure environment. In practice, sequential agent-environment interactions expose the DRL algorithms and their corresponding downstream interpretations to extra adversarial risk. In spite of the prevalence of malicious attacks, there is no existing work studying the possibility and feasibility of malicious attacks against DRL interpretations. To bridge this gap, in this paper, we investigate the vulnerability of DRL interpretation methods. Specifically, we introduce the first study of the adversarial attacks against DRL interpretations, and propose an optimization framework based on which the optimal adversarial attack strategy can be derived. In addition, we study the vulnerability of DRL interpretation methods to the model poisoning attacks, and present an algorithmic framework to rigorously formulate the proposed model poisoning attack. Finally, we conduct both theoretical analysis and extensive experiments to validate the effectiveness of the proposed malicious attacks against DRL interpretations.

The goal of metric learning is to learn a good distance metric that can capture the relationships among instances, and its importance has long been recognized in many fields. An implicit assumption in the traditional settings of metric learning is that the associated labels of the instances are deterministic. However, in many real-world applications, the associated labels come naturally with probabilities instead of deterministic values, which makes it difficult for the existing metric-learning methods to work well in these applications. To address this challenge, in this article, we study how to effectively learn the distance metric from datasets that contain probabilistic information, and then propose several novel metric-learning mechanisms for two types of probabilistic labels, i.e., the instance-wise probabilistic label and the group-wise probabilistic label. Compared with the existing metric-learning methods, our proposed mechanisms are capable of learning distance metrics directly from the probabilistic labels with high accuracy. We also theoretically analyze the proposed mechanisms and conduct extensive experiments on real-world datasets to verify the desirable properties of these mechanisms.

Treatment effect estimation refers to the estimation of causal effects, which benefits decision-making process across various domains, but it is a challenging problem in real practice. The estimation of causal effects from observational data at the individual level faces two major challenges, i.e., treatment selection bias and missing counterfactuals. Existing methods tackle the selection bias problem by learning a balanced representation and infer the missing counterfactuals based on the learned representation. However, most existing methods learn the representation in a global manner and ignore the local similarity information, which is essential for an accurate estimation of causal effects. Motivated by the above observations, we propose a novel representation learning method, which adaptively extracts fine-grained similarity information from the original feature space and minimizes the distance between different treatment groups as well as the similarity loss during the representation learning procedure. Experiments on three public datasets demonstrate that the proposed method achieves the best performance in causal effect estimation among all the compared methods and is robust to the treatment selection bias.

In many time-sensitive applications, knowing the classification results as early as possible while preserving the accuracy is extremely important for further actions. Shapelet-based early classification methods are popular due to their natural interpretability. However, most of the existing shapelet-based methods ignore the distance information between the shapelets and the time series. The distance information, though may contain some noise, can reflect more information between the shapelets and the time series. Some existing works adopt the distance information, but are not robust to the noise in the distance information. To tackle this challenge, we present a novel distance transformation based early classification (DTEC) framework, which transfers the original time series into the distance space. Upon the distance space, a probabilistic classifier is trained, and a novel classification criterion confidence area is proposed in order to overcome the noise brought by the training phase and the dataset. The effectiveness of the proposed framework is validated on three time series benchmarks as well as the extensive datasets selected from UCR time series archive.

A clinically meaningful distance metric, which is learned from measuring patient similarity, plays an important role in clinical decision support applications. Several metric learning approaches have been proposed to measure patient similarity, but they are mostly designed for learning the metric at only one time point/interval. It leads to a problem that those approaches cannot reflect the similarity variations among patients with the progression of diseases. In order to capture similarity information from multiple future time points simultaneously, we formulate a multi-task metric learning approach to identify patient similarity. However, it is challenging to directly apply traditional multi-task metric learning methods to learn such similarities due to the high dimensional, complex and noisy nature of healthcare data. Besides, the disease labels often have clinical relationships, which should not be treated as independent. Unfortunately, traditional formulation of the loss function ignores the degree of labels' similarity. To tackle the aforementioned challenges, we propose mtTSML, a multi-task triplet constrained sparse metric learning method, to monitor the similarity progression of patient pairs. In the proposed model, the distance for each task can be regarded as the combination of a common part and a task-specific one in the transformed low-rank space. We then perform sparse feature selection for each individual task to select the most discriminative information. Moreover, we use triplet constraints to guarantee the margin between similar and less similar pairs according to the ordered information of disease severity levels (i.e. labels). The experimental results on two real-world healthcare datasets show that the proposed multi-task metric learning method significantly outperforms the state-of-the-art baselines, including both single-task and multi-task metric learning methods.

Metric learning aims to learn a good distance metric that can capture the relationships among instances, and its importance has long been recognized in many fields. In the traditional settings of metric learning, an implicit assumption is that the associated labels of the instances are deterministic. However, in many real-world applications, the associated labels come naturally with probabilities instead of deterministic values. Thus, the existing metric learning methods cannot work well in these applications. To tackle this challenge, in this paper, we study how to effectively learn the distance metric from datasets that contain probabilistic information, and then propose two novel metric learning mechanisms for two types of probabilistic labels, i.e., the instance-wise probabilistic label and the group-wise probabilistic label. Compared with the existing metric learning methods, our proposed mechanisms are capable of learning distance metrics directly from the probabilistic labels with high accuracy. We also theoretically analyze the two proposed mechanisms and provide theoretical bounds on the sample complexity for both of them. Additionally, extensive experiments based on real-world datasets are conducted to verify the desirable properties of the proposed mechanisms.

Patient similarity learning aims to derive a clinically meaningful similarity metric to measure the similarity between a pair of patients according to their historical clinical information, which could help to predict the clinical outcomes of the patient of interest. However, the patient clinical data are usually complex, and contain much irrelevant and redundant information, which makes it difficult to learn the similarity metric with high accuracy. Although some methods have been proposed to address the complex nature of patient data, they overemphasize sparsity-based relevant feature selection and fail to take into consideration the redundant features that are highly correlated with each other, and this heavily degrades the accuracy of the learned results. To address the above challenges, we propose a novel uncorrelated patient similarity learning approach, which can not only select the most relevant features for the learning task, but also guarantee that the selected features have low correlations with each other. Additionally, to address the scenarios where the patient data are distributed across different sites, we extend the proposed approach and design a distributed mechanism, based on which the similarity metric can be accurately learned without directly accessing the raw patient data at each site. The desirable performance of the proposed methods are verified through extensive experiments conducted on both real-world and synthetic datasets.

As an effective way to solicit useful information from the crowd, crowdsourcing has emerged as a popular paradigm to solve challenging tasks. However, the data provided by the participating workers are not always trustworthy. In real world, there may exist malicious workers in crowdsourcing systems who conduct the data poisoning attacks for the purpose of sabotage or financial rewards. Although data aggregation methods such as majority voting are conducted on workers» labels in order to improve data quality, they are vulnerable to such attacks as they treat all the workers equally. In order to capture the variety in the reliability of workers, the Dawid-Skene model, a sophisticated data aggregation method, has been widely adopted in practice. By conducting maximum likelihood estimation (MLE) using the expectation maximization (EM) algorithm, the Dawid-Skene model can jointly estimate each worker»s reliability and conduct weighted aggregation, and thus can tolerate the data poisoning attacks to some degree. However, the Dawid-Skene model still has weakness. In this paper, we study the data poisoning attacks against such crowdsourcing systems with the Dawid-Skene model empowered. We design an intelligent attack mechanism, based on which the attacker can not only achieve maximum attack utility but also disguise the attacking behaviors. Extensive experiments based on real-world crowdsourcing datasets are conducted to verify the desirable properties of the proposed mechanism.

In this paper, we propose differentially private protocols for Naive Bayes classification over distributed data. Compared with existing works, the privacy and security models in the proposed protocols are stronger: firstly, both the miner and parties can be arbitrarily malicious and can collude with each other to violate the remaining honest parties privacy; secondly, all communication channels between them can be assumed to be insecure. Specifically, we build a guarantee of differential privacy into the cryptographic construction so that the proposed protocols can tolerate collusions and resist eavesdropping attacks which are caused by insecure communication channels. Additionally, the proposed protocols can be implemented at lower computation and communication costs, and some extensions to our protocols (e.g. supporting parties dynamic joins or leaves) are also proposed in this paper. Both theoretical analysis and simulation results show that the proposed privacy-preserving protocols for Naive Bayes have strong security and better classification performance than the standard one.