In many real-world applications, data are often collected in the form of stream, and thus the distribution usually changes in nature, which is referred as concept drift in literature. We propose a novel and effective approach to handle concept drift via model reuse, leveraging previous knowledge by reusing models. Each model is associated with a weight representing its reusability towards current data, and the weight is adaptively adjusted according to the model performance. We provide generalization and regret analysis.

Transferability captures the ability of an attack against a machine-learning model to be effective against a different, potentially unknown, model. Studying transferability of attacks has gained interest in the last years due to the deployment of cyber-attack detection services based on machine learning. For these applications of machine learning, service providers avoid disclosing information about their machine-learning algorithms. As a result, attackers trying to bypass detection are forced to craft their attacks against a surrogate model instead of the actual target model used by the service. While previous work has shown that finding test-time transferable attack samples is possible, it is not well understood how an attacker may construct adversarial examples that are likely to transfer against different models, in particular in the case of training-time poisoning attacks. In this paper, we present the first empirical analysis aimed to investigate the transferability of both test-time evasion and training-time poisoning attacks. We provide a unifying, formal definition of transferability of such attacks and show how it relates to the input gradients of the surrogate and of the target classification models. We assess to which extent some of the most well-known machine-learning systems are vulnerable to transfer attacks, and explain why such attacks succeed (or not) across different models. To this end, we leverage some interesting connections highlighted in this work among the adversarial vulnerability of machine-learning models, their regularization hyperparameters and input gradients.

Interpreting a nonparametric regression model with many predictors is known to be a challenging problem. There has been renewed interest in this topic due to the extensive use of machine learning algorithms and the difficulty in understanding and explaining their input-output relationships. This paper develops a unified framework using a derivative-based approach for existing tools in the literature, including the partial-dependence plots, marginal plots and accumulated effects plots. It proposes a new interpretation technique called the accumulated total derivative effects plot and demonstrates how its components can be used to develop extensive insights in complex regression models with correlated predictors. The techniques are illustrated through simulation results.

In this paper, we apply the NetFV and NetVLAD layers for the end-to-end language identification task. NetFV and NetVLAD layers are the differentiable implementations of the standard Fisher Vector and Vector of Locally Aggregated Descriptors (VLAD) methods, respectively. Both of them can encode a sequence of feature vectors into a fixed dimensional vector which is very important to process those variable-length utterances. We first present the relevances and differences between the classical i-vector and the aforementioned encoding schemes. Then, we construct a flexible end-to-end framework including a convolutional neural network (CNN) architecture and an encoding layer (NetFV or NetVLAD) for the language identification task. Experimental results on the NIST LRE 2007 close-set task show that the proposed system achieves significant EER reductions against the conventional i-vector baseline and the CNN temporal average pooling system, respectively.

Networks are ubiquitous structure that describes complex relationships between different entities in the real world. As a critical component of prediction task over nodes in networks, learning the feature representation of nodes has become one of the most active areas recently. Network Embedding, aiming to learn non-linear and low-dimensional feature representation based on network topology, has been proved to be helpful on tasks of network analysis, especially node classification. For many real-world systems, multiple types of relations are naturally represented by multiple networks. However, existing network embedding methods mainly focus on single network embedding and neglect the information shared among different networks. In this paper, we propose a novel multiple network embedding method based on semisupervised autoencoder, named DeepMNE, which captures complex topological structures of multi-networks and takes the correlation among multi-networks into account. We evaluate DeepMNE on the task of node classification with two real-world datasets. The experimental results demonstrate the superior performance of our method over four state-of-the-art algorithms.

Clustering is the process of finding and analyzing underlying group structures in data. In recent years, data as become increasingly higher dimensional and therefore an increased need for dimension reduction techniques for use in clustering. Although such techniques are firmly established in the literature for multivariate data, there is a relative paucity in the area of matrix variate or three way data. Furthermore, these few methods all assume matrix variate normality which is not always sensible if skewness is present. We propose a mixture of bilinear factor analyzers model using four skewed matrix variate distributions, namely the matrix variate skew-t, generalized hyperbolic, variance gamma and normal inverse Gaussian distributions.

Many questions in Data Science are fundamentally causal in that our objective is to learn the effect of some exposure (randomized or not) on an outcome interest. Even studies that are seemingly non-causal (e.g. prediction or prevalence estimation) have causal elements, such as differential censoring or measurement. As a result, we, as Data Scientists, need to consider the underlying causal mechanisms that gave rise to the data, rather than simply the pattern or association observed in the data. In this work, we review the "Causal Roadmap", a formal framework to augment our traditional statistical analyses in an effort to answer the causal questions driving our research. Specific steps of the Roadmap include clearly stating the scientific question, defining of the causal model, translating the scientific question into a causal parameter, assessing the assumptions needed to translate the causal parameter into a statistical estimand, implementation of statistical estimators including parametric and semi-parametric methods, and interpretation of our findings. Throughout we focus on the effect of an exposure occurring at a single time point and provide extensions to more advanced settings.

Anderson mixing (or Anderson acceleration) is an efficient acceleration method for fixed point iterations (i.e., $x_{t+1}=G(x_t)$), e.g., gradient descent can be viewed as iteratively applying the operation $G(x) = x-\alpha\nabla f(x)$. It is known that Anderson mixing is quite efficient in practice and can be viewed as an extension of Krylov subspace methods for nonlinear problems. First, we show that Anderson mixing with Chebyshev polynomial parameters can achieve the optimal convergence rate $O(\sqrt{\kappa}\ln\frac{1}{\epsilon})$, which improves the previous result $O(\kappa\ln\frac{1}{\epsilon})$ provided by [Toth and Kelley, 2015] for quadratic functions. Then, we provide a convergence analysis for minimizing general nonlinear problems. Besides, if the hyperparameters (e.g., the Lipschitz smooth parameter $L$) are not available, we propose a Guessing Algorithm for guessing them dynamically and also prove a similar convergence rate. Finally, the experimental results demonstrate that the proposed Anderson-Chebyshev mixing method converges significantly faster than other algorithms, e.g., vanilla gradient descent (GD), Nesterov's Accelerated GD. Also, these algorithms combined with the proposed guessing algorithm (guessing the hyperparameters dynamically) achieve much better performance.

Many machine learning, statistical inference, and portfolio optimization problems require minimization of a composition of expected value functions (CEVF). Of particular interest is the finite-sum versions of such compositional optimization problems (FS-CEVF). Compositional stochastic variance reduced gradient (C-SVRG) methods that combine stochastic compositional gradient descent (SCGD) and stochastic variance reduced gradient descent (SVRG) methods are the state-of-the-art methods for FS-CEVF problems. We introduce compositional stochastic average gradient descent (C-SAG) a novel extension of the stochastic average gradient method (SAG) to minimize composition of finite-sum functions. C-SAG, like SAG, estimates gradient by incorporating memory of previous gradient information. We present theoretical analyses of C-SAG which show that C-SAG, like SAG, and C-SVRG, achieves a linear convergence rate when the objective function is strongly convex; However, C-CAG achieves lower oracle query complexity per iteration than C-SVRG. Finally, we present results of experiments showing that C-SAG converges substantially faster than full gradient (FG), as well as C-SVRG.

This article revisits an analysis on inaccuracies of time series averaging under dynamic time warping conducted by \cite{Niennattrakul2007}. The authors presented a correctness-criterion and introduced drift-outs of averages from clusters. They claimed that averages are inaccurate if they are incorrect or drift-outs. Furthermore, they conjectured that such inaccuracies are caused by the lack of triangle inequality. We show that a rectified version of the correctness-criterion is unsatisfiable and that the concept of drift-out is geometrically and operationally inconclusive. Satisfying the triangle inequality is insufficient to achieve correctness and unnecessary to overcome the drift-out phenomenon. We place the concept of drift-out on a principled basis and show that sample means as global minimizers of a Fr\'echet function never drift out. The adjusted drift-out is a way to test to which extent an approximation is coherent. Empirical results show that solutions obtained by the state-of-the-art methods SSG and DBA are incoherent approximations of a sample mean in over a third of all trials.