The paper discusses the approach to binary image classification with ResNet based deep learning. Its actuality is the automatic image classification process and subject-the use of the methods of data augmentation and in the improvement of the model accuracy-deep neural networks. The goal of this research was to develop and assess the efficiency of using the pre-trained ResNet network for solving the binary classification problem. Such image preprocessing and implementation for the augmentation method and mixed learning were conducted to optimize the classification process. The experimental results discussed point towards integration in preprocessing methods, dynamic image loading, and computational process optimization that drive substantial growth in the quality of classification, which carries great practical value in further explorations in this area

The problem of forecasting electricity generation is key to enabling decision-making support at the level of individual prosumers in the power grid and efficient prosumer integration into the grid. This study investigates the application of machine learning (ML) approaches to photovoltaic generation forecasting, aiming to provide general practical recommendations for the development of applied forecasting solutions. To this end, a specific use case was considered in the context of a private household with a photovoltaic installation, where data was gathered for several years. Based on the experimental results, a set of recommendations for applying ML models to photovoltaic generation forecasting tasks was formulated in the context of prosumer decision support. These recommendations address key aspects such as training and test data sizes used in the model creation process, and prediction horizon size used in the prediction process. In addition, guidelines on model file size were developed from the perspective of practical model utilization in specific use cases. This research demonstrates that establishing universal guidelines for ML model utilization in the Power System (PS) domain is both beneficial and achievable. It also highlights opportunities for further research on developing solutions for automated recommendations for required training data sizes and prediction horizons

Based on the analysis of the evolution of the mass use of artificial intelligence, the direction of research related to the optimization of the information content of the modes of operation of environments with randomness is substantiated. Models of reducing the features reflecting the characteristics of the environment to a single scale are considered. For a wide class of objects with a finite or continuous number of states, represented by Kolmogorov equations, procedures for their identification are substantiated. Differential models are proposed, integrated into optimization strategies by using discrete differential spectrum, which is a reflection of the information content of the environment in the differential domain. Models are considered, integrated into strategies for choosing actions of optimization procedures

This paper implements a data-centric approach to training and testing models that assess a client’s credit rating based on their transactions that are obtained using open banking. Well-defined logic rules were used to create synthetic data, and additional synthetic datasets were generated using GANs. The training and testing of a machine learning model (Random Forest) and neural network models implemented in TensorFlow and PyTorch with identical architectures were performed. The process of adapting the models to data updates is examined. Data clustering was used to identify patterns of behaviour among clients who did not repay their loans. The quality of the generated data was evaluated. Classification results are presented depending on the type of model, the amount of data, and the consideration of client behaviour patterns. The results showed that Random Forest models provide the highest accuracy regardless of the type of dataset. The accuracy of training neural networks depends significantly on the structure of the datasets

Out-of-distribution (OOD) detection identifies samples in the test phase that are drawn from distributions distinct from that of training in-distribution (ID) samples for a trained network. According to the information bottleneck, networks that classify tabular data tend to extract labeling information from features with strong associations to ground-truth labels, discarding less relevant labeling cues. This behavior leads to a predicament in which OOD samples with limited labeling information receive high-confidence predictions, rendering the network incapable of distinguishing between ID and OOD samples. Hence, exploring more labeling information from ID samples, which makes it harder for an OOD sample to obtain high-confidence predictions, can address this over-confidence issue on tabular data. Accordingly, we propose a novel transformer chain (TC), which comprises a sequence of dependent transformers that iteratively regain discarded labeling information and integrate all the labeling information to enhance OOD detection. The generalization bound theoretically reveals that TC can balance ID generalization and OOD detection capabilities. Experimental results demonstrate that TC significantly surpasses state-of-the-art methods for OOD detection in tabular data.

The article examines the application of semi-supervised learning and computer vision methods for segmentation of spine MR images and diagnosis of deformation of intervertebral discs. Existing neural network architectures for spine MR image segmentation and semi-supervised learning methods used in medical image segmentation are reviewed. A system for diagnosis deformation of intervertebral discs is proposed, which consists of two modules: a segmentation module and a diagnostic module. The implementation of the proposed system using two convolutional neural networks is presented: U-Net for segmentation of spine MR images and ResNet for classification of the degree of deformation of each intervertebral disc based on the Pfirrmann classification of degenerative changes. The software implementation of the medical diagnosis system was developed in the Python programming language using the PyTorch library. Neural networks were trained on an open dataset of spine MR images using a variant of the Mean Teacher semi-supervised learning method. As a result of the verification, it was found that the system is capable of performing segmentation with high accuracy. It was found that the exact prediction of the grade of degenerative changes according to Pfirrmann remains a difficult task, but the introduction of another classification made it possible to increase the accuracy of diagnosis of deformation of intervertebral discs. The proposed medical system involves the addition of new diagnostic modules, which makes it possible to use it for the comprehensive analysis of various spine diseases