Brain-computer interfaces (BCIs) have undergone significant advancements in recent years. The integration of deep learning techniques, specifically transformers, has shown promising development in research and application domains. Transformers, which were originally designed for natural language processing, have now made notable inroads into BCIs, offering a unique self-attention mechanism that adeptly handles the temporal dynamics of brain signals. This comprehensive survey delves into the application of transformers in BCIs, providing readers with a lucid understanding of their foundational principles, inherent advantages, potential challenges, and diverse applications. In addition to discussing the benefits of transformers, we also address their limitations, such as computational overhead, interpretability concerns, and the data-intensive nature of these models, providing a well-rounded analysis. Furthermore, the paper sheds light on the myriad of BCI applications that have benefited from the incorporation of transformers. These applications span from motor imagery decoding, emotion recognition, and sleep stage analysis to novel ventures such as speech reconstruction. This review serves as a holistic guide for researchers and practitioners, offering a panoramic view of the transformative potential of transformers in the BCI landscape. With the inclusion of examples and references, readers will gain a deeper understanding of the topic and its significance in the field.

On-line monitoring and diagnosis of transformers have been investigated and discussed significantly in the last few decades. Vibration method is considered as one of the non-destructive and economical methods to explore transformer operating condition and evaluate transformer mechanical integrity and performance. However, transformer vibration and its evaluation criteria in transformer faulty condition are quite challenging and are not yet agreed upon. At the same time, with the advent of IoT facilities and services, it is expected that classical diagnosis techniques will be replaced with more powerful data-driven prognosis methods that can be used efficiently and effectively in smart monitoring. In this paper, we first discuss in detail an analytical approach to the transformer vibration modeling. Nevertheless, precise interpretation of transformer vibration signal through analytical models becomes unrealistic as higher harmonics are mixed with fundamental harmonics in vibration spectra. Therefore, as the next step, we aim to support the Industry 4.0 concept by utilizing the state-of-the-art machine learning and signal processing techniques to develop prognosis models of transformer operating condition based on vibration signals. Transformer turn-to-turn insulation deterioration and short circuit analysis as one the most important concerns in transformer operation is practically emulated and examined. Along with transformer short-circuit study, transformer over and under excitations are also studied and evaluated. Our constructed predictive models are able to detect transformer short-circuit fault in early stages using vibration signals before transformer catastrophic failure. Real-time information is transferred to the cloud system and results become accessible over any portable device.

Transformers are utilized in generation, transmission, and distribution power system network and face an enormous number of hazards during their course of operation. Frequency response analysis (FRA) is an inexpensive, accurate, and non-destructive technique to explore the transformer mechanical integrity very fast. However, FRA results interpretation is not being automated yet. This study introduces a new setup for FRA measurement that can assist to leave the conventional FRA data interpretation techniques and obtain smart interpretation. Hence, FRA setups and interpretation techniques are studied and formulated in this work. A new measurement technique is introduced and discussed in detail. Practical studies are performed over distribution and power transformers and FRA data are recorded for inter-disk fault. The analysis of fault severity that we achieve in this work is an advantage of the proposed measurement technique. In this regard, we employ techniques from machine learning and numerical analysis to train a predictive engine for smart interpretation of FRA data. It is revealed that the proposed intelligent technique is capable of interpreting, detecting, and classifying the transformer winding inter-disk fault and its severity. The new introduced FRA measurement setup is also able to support the online FRA data assessment.

Predicting students’ grades has emerged as a major area of investigation in education due to the desire to identify the underlying factors that influence academic performance. Because of limited success in predicting the grade point average (GPA), most of the prior research has focused on predicting grades in a specific set of classes based on students’ prior performances. The issues associated with data-driven models of GPA prediction are further amplified by a small sample size and a relatively large dimensionality of observations in an experiment. In this paper, we utilize the state-of-the-art machine learning techniques to construct and validate a predictive model of GPA solely based on a set of self-regulatory learning behaviors determined in a relatively small-sample experiment. We quantify the predictability of each constituents of the constructed model and discuss its relevance. Ultimately, the goal of grade prediction in similar experiments is to use the constructed models for the design of intervention strategies aimed at helping students at risk of academic failure. In this regard, we lay the mathematical groundwork for defining and detecting probably helpful interventions using a probabilistic predictive model of GPA. We demonstrate the application of this framework by defining basic interventions and detecting those interventions that are probably helpful to students with a low GPA. The use of self-regulatory behaviors is warranted, because the proposed interventions can be easily practiced by students.

A classifier is epistemologically vacuous without an accurate estimate of its true error rate. In situations where the number of sample points is of the same order of magnitude as the dimension of observations, serious issues arise with respect to the performance of error estimators. In this paper, we place the problem of synthesizing an error rate estimator of a common linear classifier in an asymptotic setting in which the number of sample points is kept comparable in magnitude to the dimension of observations (double asymptotic). We construct a generalized consistent estimator of the true error rate for linear discriminant analysis in the multivariate Gaussian model under the assumption of a common covariance matrix. In other words, the estimator converges to true error rate in the double asymptotic sense. We employ simulations using both synthetic and real data to compare the performance of the new estimator to the classical consistent estimator of the true error (plug-in estimator) as well as other well-known estimators. We observe that the constructed estimator can outperform other estimators of the true error in many situations in terms of bias and root-mean-square (RMS) error.