Informatica

Accurately differentiating between fungal and bacterial infections is critical for treatment but remains challenging due to slow manual diagnostics and class imbalance in medical datasets. This paper introduces OD-Abs, a one-class outlier detection classifier based on an autoencoder architecture to distinguish infection types in microscopic images. By training solely on the reference fungal class, the model learns to identify bacterial instances as outliers. Our approach yields statistically significant improvements compared to other one-class approaches, achieving an F1-score of 0.967, an AUC of 0.980, and a specificity of 0.992. Additionally, we explore the explainability of our method by analysing reconstruction error maps from both a technical and biological perspective and conclude that the model’s decision-making aligns with biologically relevant morphological features. This proof of concept study establishes the OD-Abs autoencoder model as a highly robust tool for handling the task of classifying images of bacteria and fungi from abscesses.

Limited proficiency in sign language creates communication barriers, motivating the development of robust Automatic Sign Language Recognition (SLR) systems. We address isolated SLR in a low-resource setting using few-shot metric-based meta-learning. Sign videos are encoded with spatiotemporal convolutional backbones and classified using a prototypical network, enabling generalization to unseen classes from small support sets. We compare the SlowFast architecture with state-of-the-art video models on the LSA64 benchmark under strict class-disjoint protocols. SlowFast achieves 94.33% accuracy, outperforming competing backbones and demonstrating an effective and data-efficient approach for low-resource isolated SLR.

The field of Quantum Genetic Algorithm (QGA) has seen extensive efforts to develop reasonable algorithmic designs mainly due to its potential as well as the intrinsic difficulty to reproduce the necessary processes with the limited amount of quantum resources available, although it still remains one of rather theoretical concepts of approach to achieving quantum optimization. In this paper, we propose a dynamic encoding scheme that combines quantum adaptive search with iterative approximation of the search region. The method reuses the same quantum index register while updating the classical coordinate mapping associated with its basis states, thereby increasing local coordinate resolution without globally refining the entire continuous domain. Through noiseless statevector simulations on benchmark functions, we compare the proposed method with selected QGA variants in terms of final optimization accuracy and simulated qubit usage. The results show improved optimization performance under similar maximum qubit constraints, establishing a simulation-level resource advantage.

Determining criteria importance is a crucial task in multi-criteria decision-making problems, and the Best-Worst Method (BWM) has emerged as an effective weighting technique due to its reduced number of pairwise comparisons. Although BWM has been extended to handle group decision-making and linguistic information, existing approaches do not adequately address disagreements among evaluators, which may lead to dissatisfaction with the resulting criteria weights. To address this limitation, this paper proposes a consensus-based BWM for linguistic multi-criteria group decision-making. The proposed approach obtains consensual collective weights by minimally modifying the evaluators’ initial linguistic preferences. The resulting solution can support moderators and evaluators during consensus-reaching processes by facilitating the identification of disagreements and the generation of appropriate recommendations. The proposal is developed within the 2-tuple linguistic framework and provides both numerical and linguistic representations of the resulting weights. In addition, a novel stochastic consistency index is introduced to assess the reliability of evaluators’ preferences. The feasibility and effectiveness of the proposal are illustrated through a real-world case study and analysed by means of comparative and sensitivity analyses.

Propaganda techniques are a key tool for creating misleading content, often disseminated in native languages to increase their impact. Therefore, it is increasingly important to develop detection models not only for high-resource languages but also for low-resource languages, which still face significant limitations in propaganda detection. This study presents the first approach to automated propaganda technique detection in Lithuanian using the HALT-PROP corpus. We adapt the standard framework to account for frequent overlap between techniques. Experiments with the Lithuanian transformer LT-MLKM-modernBERT show that BILOU tagging improves span identification, while sentence classification based on span-level information enhances technique detection for most techniques. The results also indicate that training separate binary classifiers is more effective than multi-label classification in this setting. Overall, the proposed approach outperforms GPT-5.3 on most techniques and provides a strong baseline for propaganda technique detection in Lithuanian.

Shadowed set (SHS) theory was introduced by Witold Pedrycz in 1998 to play a vital role in handling vagueness and providing a way to approximate a given fuzzy set (FS) by a construct, which is easy to deal with in practice. Nonetheless, the application of SHSs and shadowed numbers (SHNs) is surprisingly underrepresented in fuzzy multi-attribute decision-making (FMADM) literature. To address this gap, the current research aims to introduce and demonstrate a new FMADM method called the SHAdowed $\boldsymbol{N}\boldsymbol{U}$mber based $\boldsymbol{R}\mathit{anking}\hspace{2.5pt}\mathit{and}\boldsymbol{S}\mathit{election}\hspace{2.5pt}\boldsymbol{A}\mathit{pproach}$ (SHANURSA). The methodology involves, inter alia, transforming FNs into their corresponding SHNs using Grzegorzewski’s (2013) approach, then implementing an efficient weighted Minkowski Distance Metric (MDM) for SHNs to better gauge the proximity/remoteness between them. Additionally, a new and unusual closeness coefficient is introduced and defined to enhance the ranking accuracy of alternatives. Finally, the applicability and effectiveness of this closeness coefficient and proximity-based FMADM method is demonstrated by means of three distinct case studies from literature. The findings of this research show that the suggested method is trustworthy.

A new population-based metaheuristic called Zebrafish Optimization Algorithm (ZFO) is proposed to find the global optimum solution using foraging behaviour of zebrafish larvae (Danio rerio). In ZFO, an exploration and exploitation trade-off is realized by simulation of exploratory dispersion, collaborative shoaling and directional searching. The movement strategy incorporates a stagnation-avoiding reinitialization mechanism, ensuring that diversity is maintained and premature convergence is avoided. To demonstrate the efficiency of ZFO, 23 benchmark functions including unimodal, multimodal and fixed dimension multimodal functions are tested and their optimal solutions are searched. Results obtained by ZFO are compared with existing metaheuristic algorithms, namely Cuttlefish Optimization, Jellyfish Search, Pufferfish Optimizer and Krill Herd algorithm. The Friedman rank test and Wilcoxon signed-rank test show that ZFO yields better-quality solutions than the compared algorithms in terms of convergence speed and robustness. The novel characteristic of ZFO is that it combines biologically inspired zebrafish dynamics and adaptive diversification strategy to enhance the capability of global search.

Considering the landscape information of the optimization problem represents a promising research direction, as landscape features can provide a metaheuristic algorithm with useful information for understanding the search state, allowing it to adjust the search strategy accordingly. As a successful branch of metaheuristic algorithms for solving single-objective bound-constrained problems, however, the original design of differential evolution (DE) does not sufficiently consider information about the landscape of the solution space. To address this, an effective DE based on reinforcement learning (RL) that dynamically selects a suitable mutation operator via landscape information during the convergence process is proposed in this study, which consists of two phases: offline and online. In the “offline” phase, the proposed algorithm uses an RL-based algorithm to construct a learning model to understand the relationships between landscape characteristics and search operators. Then, in the “online” phase, the learning model constructed from the offline phase and a lightweight fitness landscape analysis (FLA) method are used by the DE to dynamically determine the suitable mutation operator based on the search state encountered by DE during the convergence process, which can reduce the costs for the FLA in every iteration. To understand the performance of the proposed algorithm, the CEC2021 and CEC2022 benchmark functions are used to evaluate its search performance against different DE-based algorithms for solving single-objective optimization problems. Simulation results show that the proposed algorithm outperforms other state-of-the-art DE-based algorithms and other DE algorithms based on FLA and RL in most cases.

This paper firstly distinguishes between risk and Risk Factor (RF), as fundamental concepts in risk management. The need for the current study lies in the fact that although there are various risk analysis methods, but there is no well-defined frameworks for RF analysis. Manifestly, most of the researches on RF have concentrated on identifying them, expressing relationships among them, and descriptive discussion on their root causes, rather than numerically analysing them. Additionally, most of the risk-related methods (e.g. the classical Failure Mode and Effects Analysis (FMEA)) are very sensitive to nuanced changes on inputs. Thus, the paper is to design a robust RF analysis method. The paper proposes a new method called Risk Influential Factor Analysis (RIFA) for evaluation of RFs. Additionally, for getting reliable results, weighting RFs is done by a novel idea named Weight Estimation by Steady Trend (WEST). The WEST is established on the basis of a well-known branch of Multiple Attribute Decision-Making (MADM), called surrogate weighting. Therewith, in analysis section, the paper shows that (I) employing the suggested hybrid RIFA-WEST methodology leads to robust results, (II) compared to the classical FMEA, the RIFA has not many drawbacks to it, and (III) the WEST is comparable with many of the existing surrogate weighting methods. The proposed methodology is applied in a case study from one of the upstream oil engineering disciplines, i.e. Chemical Enhanced Oil Recovery (CEOR).