Institute of Computer Science

Quantitative trading strategies rely on accurately ranking stocks to identify profitable investments. Effective portfolio management requires models that can reliably order future stock returns. Transformer models are promising for understanding financial time series, but how different training loss functions affect their ability to rank stocks well is not yet fully understood. Financial markets are challenging due to their changing nature and complex relationships between stocks. Standard loss functions, which aim for simple prediction accuracy, often aren't enough. They don't directly teach models to learn the correct order of stock returns. While many advanced ranking losses exist from fields such as information retrieval, there hasn't been a thorough comparison to see how well they work for ranking financial returns, especially when used with modern Transformer models for stock selection. This paper addresses this gap by systematically evaluating a diverse set of advanced loss functions including pointwise, pairwise, listwise for daily stock return forecasting to facilitate rank-based portfolio selection on S&P 500 data. We focus on assessing how each loss function influences the model's ability to discern profitable relative orderings among assets. Our research contributes a comprehensive benchmark revealing how different loss functions impact a model's ability to learn cross-sectional and temporal patterns crucial for portfolio selection, thereby offering practical guidance for optimizing ranking-based trading strategies.

This article presents an architecture for engineering reusable embedded software using modern C++ principles and a custom–built dependency injection framework. It details the framework’s design, specifically tailored for resource-constrained environments. The framework promotes modular and testable architecture. Its data–driven (via Json file) configuration defines component dependencies and determines their instantiation. The article demonstrates how such approach facilitates component decoupling and provides a viable path for developers to create scalable, portable, and high-quality embedded software, significantly reducing future development efforts.

Modern scientific knowledge management systems increasingly demand methods capable of flexible modeling and dynamic analytics. This paper presents an approach to modeling for the needs of research knowledge base systems. The presented method integrates extensible XML-based data modeling with advanced analytical tools. The approach enables the structured definition of domain models characterized by hierarchical nesting, historical tracking, and semantic versioning. The presented analytical engine, based on this approach, uses XPath navigation to support dynamic pivot-table aggregations over complex, nested data. The underlying modeling is particularly useful for handling temporal data. It is suitable for broader contexts, resulting in flexible data structures, multi-version record management, interoperability with Linked Open Data (LOD) standards, and FAIR-compliant workflows. The results demonstrate that combining model-centric extensibility, semantic interoperability, and user-driven analytics provides a scalable and adaptable foundation for building information management systems across diverse institutional and organizational settings. A use case for a university knowledge database is presented, and acceptance of the system by users is discussed.

We analyze long-range dependence (LRD) for word time series, understood as a slower than exponential decay of the two-point Shannon mutual information. We achieve this by examining the decay of the cosine correlation, a proxy object defined in terms of the cosine similarity between word2vec embeddings of two words, computed by an analogy to the Pearson correlation. By the Pinsker inequality, the squared cosine correlation between two random vectors lower bounds the mutual information between them. Using the Standardized Project Gutenberg Corpus, we find that the cosine correlation between word2vec embeddings exhibits a readily visible stretched exponential decay for lags roughly up to 1000 words, thus corroborating the presence of LRD. By contrast, for the Human vs. LLM Text Corpus entailing texts generated by large language models, there is no systematic signal of LRD. Our findings may support the need for novel memory-rich architectures in large language models that exceed not only hidden Markov models but also Transformers.

This study explores the potential of using convolutional neural networks (CNNs) to diagnose dementia early and manage it in an individualized way. Segmented brain magnetic resonance imaging (MRI) images from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database represented Alzheimer’s disease (AD), mild cognitive impairment (MCI), and cognitively normal (CN) subjects. These classes served to train, validate, and test CNN-based models. The first four models were developed entirely from scratch, and the other four employed transfer learning (TL). While both approaches demonstrated high classification accuracy (93.69% on average), TL-based models outperformed independently developed ones, achieving 97.64% accuracy compared with 89.75%. The CNN-based models yielded information about detected dementia type, diagnosis confidence level, and gradient-weighted class activation mapping (Grad-CAM)-generated heatmaps highlighting pathologically affected brain regions. These results indicate the high potential of CNN-based models for enhancing early dementia detection and differentiation and offer a promising basis for developing deep learning (DL)-based clinical decision support systems (CDSSs). Such systems could assist healthcare professionals in reducing dementia diagnosis time, optimizing patient-tailored management and treatment strategies, and improving the quality of life for individuals with dementia.

Fourier transform infrared spectroscopy (FTIR) was used to investigate chemical differences in prostate tissue caused by prostate cancer and to correlate these data with medical. In FTIR spectra of prostate tissues, a higher amount of peaks originating from phospholipids, amide, and lipid vibrations was detected in comparison with FTIR spectra of control prostate tissues. Principal Component Analysis (PCA) showed that it is possible to differentiate two types of tissues using FTIR range corresponding to (i) phospholipids and amides and (ii) lipids. Machine learning methods showed that values of area under the curve (AUC), accuracy, F1, precision, and recall were higher for the fingerprint range than for the second one. However, values of all these parameters in both analyzed ranges were higher than 0.95. Moreover, the proposed FTIR marker of prostate cancer (wavenumber at 1685 cm-1) correlated with bGleason, pGleason, and ISUP, which suggested that FTIR spectroscopy reflected the medical characterization of prostate cancer.

In many binary classification applications, the costs of false positives and negatives are imbalanced. Furthermore, there is often uncertainty about the exact costs of these errors. A natural measure-of-interest to be minimised in such scenarios is the expected misclassification cost. We identify many situations where this measure has analytic gradients, and thus it can be used as a training loss and optimised directly using empirical risk minimisation. In particular, we derive such losses from the Beta, Gamma and Gaussian distributions to model different kinds of cost uncertainty. The Beta family includes commonly used losses such as cross-entropy, squared error and 0–1 loss as special cases. The question then arises as to when it is appropriate to directly optimize the measure-of-interest, versus using a standard surrogate like cross-entropy or focal loss during training. After revisiting the theory of surrogate losses, proper losses and cost-sensitive learning to obtain good candidate surrogates out of derived families, we conduct an empirical comparison of derived training losses that, to our knowledge, were never tried on deep neural networks before, with the aim to minimise cost-sensitive measures-of-interest. The findings show that using Beta losses in training leads to improved performance compared to traditional training objectives like cross-entropy, label smoothing, and focal loss. This improvement is seen not only in terms of misclassification cost metrics, but (perhaps surprisingly) also in conventional metrics such as accuracy, mean squared error, and the area under the ROC curve.

NaviSecure is an indoor navigation system developed atUniversity of Siedlce. It utilizes a dedicated hierarchical building map and an infrastructure of Bluetooth Low Energy transmitters, as well as hazard detectors such as smoke, flood, and gas sensors. The system facilitates daily navigation and provides emergency assistance while addressing the needs of individuals with disabilities. In this paper we present the pathfinding module of the NaviSecure system, implemented on top of the Neo4j graph database. We introduce the basic concepts behind our approach, we discuss the system architecture and our custom approach to deal with special needs of the users. Finally, we show the results of our experimental evaluation of several graph algorithms, modified to meet NaviSecure’s specific requirements. The results confirm the efficiency of our approach.