Many real-world applications, e.g., healthcare, present multi-variate time series prediction problems. In such settings, in addition to the predictive accuracy of the models, model transparency and explainability are paramount. We consider the problem of building explainable classifiers from multi-variate time series data. A key criterion to understand such predictive models involves elucidating and quantifying the contribution of time varying input variables to the classification. Hence, we introduce a novel, modular, convolution-based feature extraction and attention mechanism that simultaneously identifies the variables as well as time intervals which determine the classifier output. We present results of extensive experiments with several benchmark data sets that show that the proposed method outperforms the state-of-the-art baseline methods on multi-variate time series classification task. The results of our case studies demonstrate that the variables and time intervals identified by the proposed method make sense relative to available domain knowledge.

Many applications, e.g., healthcare, education, call for effective methods methods for constructing predictive models from high dimensional time series data where the relationship between variables can be complex and vary over time. In such settings, the underlying system undergoes a sequence of unobserved transitions among a finite set of hidden states. Furthermore, the relationships between the observed variables and their temporal dynamics may depend on the hidden state of the system. To further complicate matters, the hidden state sequences underlying the observed data from different individuals may not be aligned relative to a common frame of reference. Against this background, we consider the novel problem of jointly learning the state-dependent inter-variable relationships as well as the pattern of transitions between hidden states from multi-variate time series data. To solve this problem, we introduce the State-Regularized Vector Autoregressive Model (SrVARM) which combines a state-regularized recurrent neural network to learn the dynamics of transitions between discrete hidden states with an augmented autoregressive model which models the inter-variable dependencies in each state using a state-dependent directed acyclic graph (DAG). We propose an efficient algorithm for training SrVARM by leveraging a recently introduced reformulation of the combinatorial problem of optimizing the DAG structure with respect to a scoring function into a continuous optimization problem. We report results of extensive experiments with simulated data as well as a real-world benchmark that show that SrVARM outperforms state-of-the-art baselines in recovering the unobserved state transitions and discovering the state-dependent relationships among variables.

Graph Neural Networks (GNN) offer the powerful approach to node classification in complex networks across many domains including social media, E-commerce, and FinTech. However, recent studies show that GNNs are vulnerable to attacks aimed at adversely impacting their node classification performance. Existing studies of adversarial attacks on GNN focus primarily on manipulating the connectivity between existing nodes, a task that requires greater effort on the part of the attacker in real-world applications. In contrast, it is much more expedient on the part of the attacker to inject adversarial nodes, e.g., fake profiles with forged links, into existing graphs so as to reduce the performance of the GNN in classifying existing nodes.

Data from many real-world applications can be naturally represented by multi-view networks where the different views encode different types of relationships (e.g., friendship, shared interests in music, etc.) between real-world individuals or entities. There is an urgent need for methods to obtain low-dimensional, information preserving and typically nonlinear embeddings of such multi-view networks. However, most of the work on multi-view learning focuses on data that lack a network structure, and most of the work on network embeddings has focused primarily on single-view networks. Against this background, we consider the multi-view network representation learning problem, i.e., the problem of constructing low-dimensional information preserving embeddings of multi-view networks. Specifically, we investigate a novel Generative Adversarial Network (GAN) framework for Multi-View Network Embedding, namely MEGAN, aimed at preserving the information from the individual network views, while accounting for connectivity across (and hence complementarity of and correlations between) different views. The results of our experiments on two real-world multi-view data sets show that the embeddings obtained using MEGAN outperform the state-of-the-art methods on node classification, link prediction and visualization tasks.

The class imbalance problem in machine learning occurs when certain classes are underrepresented relative to the others, leading to a learning bias toward the majority classes. To cope with the skewed class distribution, many learning methods featuring minority oversampling have been proposed, which are proved to be effective. To reduce information loss during feature space projection, this study proposes a novel oversampling algorithm, named minority oversampling in kernel adaptive subspaces (MOKAS), which exploits the invariant feature extraction capability of a kernel version of the adaptive subspace self-organizing maps. The synthetic instances are generated from well-trained subspaces and then their pre-images are reconstructed in the input space. Additionally, these instances characterize nonlinear structures present in the minority class data distribution and help the learning algorithms to counterbalance the skewed class distribution in a desirable manner. Experimental results on both real and synthetic data show that the proposed MOKAS is capable of modeling complex data distribution and outperforms a set of state-of-the-art oversampling algorithms.

A brain-computer interface (BCI) system using elec-troencephalography signals provides a convenient means of communication between the human brain and a computer. Motor imagery (MI), in which motor actions are mentally rehearsed without engaging in actual physical execution, has been widely used as a major BCI approach. One robust algorithm that can successfully cope with the individual differences in MI-related rhythmic patterns is to create diverse ensemble classifiers using the subband common spatial pattern (SBCSP) method. To aggregate outputs of ensemble members, this study uses fuzzy integral with particle swarm optimization (PSO), which can regulate subject-specific parameters for the assignment of optimal confidence levels for classifiers. The proposed system combining SBCSP, fuzzy integral, and PSO exhibits robust performance for offline single-trial classification of MI and real-time control of a robotic arm using MI. This paper represents the first attempt to utilize fuzzy fusion technique to attack the individual differences problem of MI applications in real-world noisy environments. The results of this study demonstrate the practical feasibility of implementing the proposed method for real-world applications.

In this paper, we develop a novel multi-fusion brain-computer interface (BCI) based on linear discriminant analysis (LDA) to deal with motor imagery (MI) classification problem. We combine filter bank and sub-band common spatial pattern (SBCSP) to extract features from EEG data in the preprocessing phase, and then LDA classifiers are applied to classify brain activities to identify either left or right hand imagery. To further foster the performance of the proposed system, a fuzzy integral (FI) approach is employed to fuse information sources, and particle swarm optimization (PSO) algorithm is exploited to globally update parameters in the fusion structure. Consequently, our experimental results indicate that the proposed system provides superior performance compared to other approaches.

In this study, we develop a novel multi-fusion brain-computer interface (BCI) system based on a fuzzy neural network (FNN) and information fusion approaches to cope with a classification task for identifying right/left hand motor imagery. In the proposed system, we utilize a filter bank and sub-band common spatial pattern (SBCSP) to extract features from raw EEG data. A self-organizing neural fuzzy inference network (SONFIN) is then applied for a recognition task. In order to improve the classification performance, we form a committee of networks and employ fuzzy integral (FI) to attain a joint decision. To further optimize the fusion approaches, a particle swarm optimization (PSO) algorithm is exploited to globally update parameters used in the fusion stage. In consequence, our experimental result shows that the proposed fuzzy fusion system possesses superior performance compared to other comparative models.

This study proposes an EEG-based forecasting system based on a functional-link recurrent self-evolving fuzzy neural network (FL-RSEFNN) for assessing mental fatigue during a highway driving task. Drivers' cognitive states significantly affect driving safety, especially for fatigue or drowsy driving which is one of common factors to endanger individuals and the public safety. In this study, a FL-RSEFNN employs an on-line gradient descent (GD) learning rule to address the EEG regression problem in brain dynamics for estimation of driving fatigue. We analyze brain dynamics in a car driving task, which is constructed in a simulated virtual reality (VR) environment. The EEG-based forecasting system is evaluated using the generalized cross-subject approach, and the results indicate that the FLRSEFNN is superior to state-of-the-art models regardless of the use of recurrent or non-recurrent structures.

An ensemble empirical mode decomposition (EEMD)-based approach was developed to extract steady-state visual evoked potentials (SSVEPs) for wireless handling of a small robot car. Three visual stimuli, flickering at 13, 14, and 15 Hz, were displayed on a liquid crystal display monitor to induce user's SSVEPs. The induced SSVEPs were used to control three movement functions (forward, left, and right) of the small robot car. Users gazed at one chosen visual stimulus at one time, and the induced SSVEP was recognized to activate the desired movement function. In this paper, all subjects were requested to handle the small robot car to complete an S-shaped course four times. The proposed system utilized only one electroencephalography (EEG) channel placed at the Oz position. The acquired EEG signals were first segmented into 1-s epochs, and each epoch was then decomposed by EEMD into a series of oscillation components, denoted as intrinsic oscillatory functions (IOFs), representing multiscale features of the recorded signal. The SSVEP-related IOFs were then recognized using a matched filter detector (MFD), including a matched filter demodulator and an amplitude detector. The visual stimulus, which contributed maximum power to the MFD, was recognized as the gazed target. In this paper, all subjects could actuate the small robot car using the proposed EEMD-based brain computer interface system to complete an S-shaped course four times; the mean execution time, number of valid detections, and command transfer interval over the 11 subjects were 84.5 s, 51.13 commands, and 1.65 s/command, respectively.