This study explores the feasibility of developing a dynamic cognovisor capable of recognizing cognitive states and transitions using fMRI data. Data were collected from 31 participants performing spatial and verbal tasks during fMRI scanning and were preprocessed using a nine-step algorithm for artifact removal and denoising. Three types of classification problems were examined, with machine learning methods and dimensionality reduction techniques applied to classify activity states. The best-performing models were identified for each classification problem, providing insights into their applicability. Notably, binary classification of resting versus active states achieved good quality with relatively simple methods. A key finding underscores the importance of accounting for temporal history of the signal prior to the prediction moment to improve model performance.

The paper explores the possibilities of using data classification methods when forecasting time series of the geomagnetic Kp-index by machine learning methods. To classify categories of the Kp-index based on the degree of disturbance, linear and logistic regression, random forest, gradient boosting on top of decision trees, and artificial neural networks of various architectures are used. The results of these methods are compared with a trivial inertial forecast (the statistical indicators of which for problems of this type are always high) at horizons from 3 h to 1 day in 3-h increments. The problem of choosing a cross-validation scheme for selecting the model hyperparameters, ways to overcome the imbalance of categories, the relative importance of input features, as well as the dependence of the results on the test sample (beginning of the 25th solar activity cycle) on inclusion in the training sample of data from the 23rd and 24th cycles or only the 24th cycles are studied. Based on the results, conclusions are drawn about the preferred methods for classifying values of the Kp-index based on the level of geomagnetic disturbance. Ways for further research and possible improvement of the classification quality are outlined, including for determining the characteristic hidden states of Earth’s magnetosphere as a dynamic system in order to improve the quality of forecasting geomagnetic indices.

There are a lot of studies researching automated recognition of emotions. Emotions are represented as points in an emotion space. The emotion space itself is represented by different types of models. One is Facial Action Units System, another is Valence-Arousal-Dominance model. This study aims to create a mapping between these two emotion spaces. The data for the study was collected in a series of experiments with real humans, where both types of measurements were collected simultaneously. Given the data, we study the ability of machine learning models to create this type of mapping. We test different types of models against the task, such as tree-based models and linear models, and make conclusions about the optimal model.

Among dimensional models of emotions, two- and three-dimensional are most popular, while the true dimension of affective space is a matter of debates. Here we study the inherent dimension of the emotion space represented in facial expressions, along with the mapping of electromyography (EMG) signals recorded from facial muscles to expressed emotions. For this purpose, an experiment was conducted with parallel EMG recording from three facial muscles (Zygomaticus Major, Corrugator Supercilii, and Masseter) and video registration of the face with automated emotion recognition from the video stream. Data analysis based on machine learning methods confirmed the 3D nature of the affective space (at least its part reflected in facial expressions). This result is consistent with the VAD and PAD models. Possibilities of accounting for complex, higher-order, or social emotions without introducing additional dimensions are discussed. The second finding of this study is the ability to reconstruct all three significant principal components of expressed affects using EMG signals recorded from three facial muscles with the help of machine learning.

Magnetic storms can cause disruptions in the operation of radio communications, pipelines, power lines, and electrical networks, and they may possibly cause human health problems. Therefore, prediction of geomagnetic disturbances is of great practical value. Geomagnetic disturbances are usually described with the help of geomagnetic indices, including the planetary index K_p which is provided at a 3-h interval. The approach used in this study implies classifying geomagnetic disturbances according to the level of the K_p index. To do so, the whole range of the index values is divided into several intervals according to the degree of disturbance. The input data are time series of parameters of solar wind and interplanetary magnetic field, measured onboard spacecraft at the L1 Lagrange point between the Sun and the Earth, aa well as the value of the K_p index itself. To account for the ‘‘memory’’ of the time series, delay embedding of all the parameters is used—for each of the parameters, its several preceding values are taken into account. Additional preprocessing of the parameters is performed by calculating moving averages and other statistical indicators of the time series. To perform classification, various machine learning methods such as gradient boosting and artificial neural networks are used. The optimal values of the parameters of each method are determined by cross-validation, and pattern misbalance among the classes is partially reduced using the SMOTE technique. It is demonstrated that the suggested approach outperforms the trivial inertial model for all the values of the prediction horizon from 3 to 24 h (with a 3-h step). The most efficient preprocessing methods are described, as well as the best machine learning models.

In the previous studies, the integration of optical spectroscopy methods was investigated in order to increase the accuracy of the solution obtained by machine learning methods. The joint use of Raman spectroscopy and optical absorption spectroscopy to determine the concentration of heavy metal ions in water by artificial neural networks was considered. Direct training of neural networks on the data of both types of spectroscopy did not allow us to improve the result in comparison with the individual use of absorption spectroscopy data. In this study, we consider the adaptation of transfer learning approach to the integration of optical spectroscopy methods, which consists in initial training of the neural networks on the data of only the weaker method (Raman spectroscopy), followed by additional training on the data of two methods (Raman and absorption spectroscopy).

The report presents an algorithm for constructing a convolutional hierarchical neural network classifier, which is a modification of the algorithm for constructing hierarchical neural network classifiers suggested before. The original algorithm was designed to exploit intrinsic class hierarchy to build a class tree with a neural network in each node classifying groups of initial classes (in a non-terminal node) or a subset of original classes (in a terminal node). The convolutional modification utilizes convolutional neural networks instead of regular fully connected networks in order to apply the model to image classification tasks. Use of class hierarchy for image classification should reduce the number of adjusted neural network parameters compared to deep convolutional neural networks, and therefore it should reduce training and inference time. In this context the algorithm may be compared with some pruning techniques. The convolutional hierarchical neural network classifier inherits some hyperparameters of a conventional hierarchical neural network classifier, like the activation threshold and the threshold by the share of voting patterns. The goal of this study was to explore different strategies of choosing these hyperparameters. To test these strategies, we used the CIFAR-10 dataset. Also, for demonstration purposes we apply the convolutional hierarchical neural network classifier to the CIFAR-100 dataset.

The paper presents an algorithm for constructing convolutional hierarchical neural network classifiers designed for image classification. It is a modification of the algorithm for constructing hierarchical neural network classifiers. The algorithm is tested on the CIFAR-10 problem.