Iskanje po repozitoriju

A+ | A- | | SLO | ENG

Povzetek: This study deploys a scalable machine learning pipeline: combining a transformer-based classifier applied to 2.01 million English-language AI-related news headlines (July 2022–July 2024) with large-language-model and human-annotator validation (three annotators, Fleiss’ �=0.80 ) on stratified subsamples, to extract six interpretable, bias-linked discourse indicators computed at the AI-domain level: evaluative orientation (valence), loss salience, narrative drift, exposure-adjusted sentiment, cross-source divergence, and novelty-phase framing. Each operationalizes an established cognitive-psychology construct as a computable property of the information environment associated with biased risk–benefit reasoning. Results show systematic variation across domains: technical and methodological areas such as deep learning and natural language processing exhibit gain-salient framing, while safety-critical topics such as deepfakes (loss-to-gain headline ratio = 3.17) and facial recognition show strongly loss-salient profiles. Cross-model validation using an LLM on a stratified sample of 1000 headlines confirms that domain-level indicator rankings are robust to classifier choice (Spearman �=0.83 ; �<0.001 ), establishing the rank stability of pipeline outputs independently of the specific classification architecture. As a contextual application, domain-level profiles are mapped to European Union AI governance instruments, documenting parallels between discourse patterns and regulatory risk tiers. The framework provides a scalable, reproducible methodology for monitoring evaluative conditions in technology news across domains, sources, and time.
Ključne besede: applied machine learning, large scale text classification, discourse indicators, transformer-based sentiment analysis, AI-related new headlines, domain-resolved analysis, cross-model validation, large-language models
Objavljeno v DiRROS: 01.09.2026; Ogledov: 90; Prenosov: 50
Celotno besedilo (1,85 MB)
Gradivo ima več datotek! Več...

Povzetek: Eyes open and eyes closed data is often used to validate novel human brain activity classification methods. The cross-validation of models trained on minimally preprocessed data is frequently utilized, regardless of electroencephalography data comprised of data resulting from muscle activity and environmental noise, affecting classification accuracy. Moreover, electroencephalography data of a single subject is often divided into smaller parts, due to limited availability of large datasets. The most frequently used method for model validation is cross-validation, even though the results may be affected by overfitting to the specifics of brain activity of limited subjects. To test the effects of preprocessing and classifier validation on classification accuracy, we tested fourteen classification algorithms implemented in WEKA and MATLAB, tested on comprehensively and simply preprocessed electroencephalography data. Hold-out and cross-validation were used to compare the classification accuracy of eyes open and closed data. The data of 50 subjects, with four minutes of data with eyes closed and open each was used. The algorithms trained on simply preprocessed data were superior to the ones trained on comprehensively preprocessed data in cross-validation testing. The reverse was true when hold-out accuracy was examined. Significant increases in hold-out accuracy were observed if the data of different subjects was not strictly separated between the test and training datasets, showing the presence of overfitting. The results show that comprehensive data preprocessing can be advantageous for subject invariant classification, while higher subject-specific accuracy can be attained with simple preprocessing. Researchers should thus state the final intended use of their classifier.
Ključne besede: electroencephalography (EEG), machine learning, model validation
Objavljeno v DiRROS: 04.08.2026; Ogledov: 233; Prenosov: 135
Celotno besedilo (5,68 MB)
Gradivo ima več datotek! Več...

Povzetek: Dynamic Energy Budget (DEB) theory provides a coherent framework to describe how organisms acquire and use energy and matter throughout the life cycle in response to environmental conditions. The wide applicability suggests the generality of DEB theory and is one of its main strengths. However, this same generality introduces a substantial degree of abstraction, which can make model development challenging, particularly for newcomers. This paper aims to support more transparent, biologically grounded, and reproducible DEB modeling by proposing guidelines to make the development process more explicit and accessible. We identify three main challenges: (1) how to link biological processes to their mathematical representation, (2) which observations are needed to inform key physiological processes, and (3) how to carry out parameter estimation and model validation in a transparent and reproducible way. To help overcome these issues, we briefly introduce core concepts and discuss the six main steps in the DEB modeling workflow: model design, data collection and assessment, implementation, calibration, validation, and publication. These steps are illustrated with a case study on zebrafish (Danio rerio) that demonstrates how data availability influences model development and the range of predictions that can be supported. By proposing a structured workflow, we hope to encourage researchers to explore the full potential of DEB theory.
Ključne besede: modeling development workflow, bioenergetics, good modeling practices, estimation in-context, model calibration, model validation, add-my-pet
Objavljeno v DiRROS: 08.07.2026; Ogledov: 251; Prenosov: 270
Celotno besedilo (2,89 MB)
Gradivo ima več datotek! Več...

Povzetek: Background: The antiepileptic drug lamotrigine (LTG) shows high pharmacokinetic variability due to genotype influence and concomitant use of glucuronidation inducers and inhibitors, both of which may be frequently taken by elderly patients. Our goal was to develop a reliable quantification method for lamotrigine and its main glucuronide metabolite lamotrigine-N2-glucuronide (LTG-N2-GLU) in dried blood spots (DBS) to enable routine therapeutic drug monitoring and to identify altered metabolic activity for early detection of drug interactions possibly leading to suboptimal drug response. Results: The analytical method was validated in terms of selectivity, accuracy, precision, matrix effects, haematocrit, blood spot volume influence, and stability. It was applied to a clinical study, and the DBS results were compared to the concentrations determined in plasma samples. A good correlation was established for both analytes in DBS and plasma samples, taking into account the haematocrit and blood cell-to-plasma partition coefficients. It was demonstrated that the method is suitable for the determination of the metabolite-to-parent ratio to reveal the metabolic status of individual patients. Conclusions: The clinical validation performed confirmed that the DBS technique is a reliable alternative for plasma lamotrigine and its glucuronide determination.
Ključne besede: dried blood spot, lamotrigine, lamotrigine glucuronide, therapeutic drug monitoring, clinical validation, haematocrit effect
Objavljeno v DiRROS: 16.06.2026; Ogledov: 272; Prenosov: 193
Celotno besedilo (2,87 MB)
Gradivo ima več datotek! Več...

Povzetek: Human papillomavirus (HPV)-based screening offers better protection against cervical cancer compared to cytology, but HPV screening assays must adhere to validation requirements of the international guidelines to ensure optimal performance. Allplex HPV HR Detection (Allplex) assay, launched in the late 2022, is a fully automated real-time PCR-based assay utilizing innovative technology that enables quantification and concurrent distinction of 14 high-risk HPV genotypes (HPV16,18,31,33,35,39,45,51,52,56,58,59,66 and 68). We assessed the validity of the Allplex for cervical cancer screening purposes, via comparison to a clinically validated comparator assay (Hybrid Capture 2; HC2), and through assessment of intra-laboratory reproducibility and inter-laboratory agreement. A clinical validation panel comprised of 973 residual ThinPrep samples was obtained from women aged 30-64 years participating in the organized Slovenian screening program, of these 863 were from women undergoing their regular screening visit after a previous negative screen test while 110 were from women with underlying cervical intraepithelial neoplasia grade 2 or worse (CIN2+) lesions. The Allplex's relative clinical sensitivity for detection of CIN2+ and CIN3+ were 1.01 (95%CI;0.98-1.04) and 0.98 (95%CI;0.95-1.02), compared to that of HC2. At recommended thresholds of ≥98% and ≥90%, the Allplex's clinical sensitivity and specificity (p=0.0004 and p=0.02, respectively) were non-inferior to HC2. High intra-laboratory reproducibility and inter-laboratory agreement, both overall (98.1% and 97.9%, respectively) and at genotype level (>98.7%) was observed. In addition, analytical genotype-specific performance of Allplex was compared to that of its predecessor Anyplex HPV HR; high overall agreement was observed (96.3%; kappa value 0.88), with some variations in performance. In conclusion, Allplex met all validation criteria described in the international guidelines on sensitivity, specificity and laboratory reproducibility and can be considered clinically validated for primary cervical cancer screening.
Ključne besede: Allplex HPV HR, HPV, cervical cancer, genotyping, human papillomaviruses, screening, validation
Objavljeno v DiRROS: 11.06.2026; Ogledov: 303; Prenosov: 261
Celotno besedilo (443,83 KB)
Gradivo ima več datotek! Več...