Driven by significant improvements in architectural design and training pipelines, computer vision has recently experienced dramatic progress in terms of accuracy on classic benchmarks such as ImageNet. These highly-accurate models are challenging to deploy, as they appear harder to compress using standard techniques such as pruning. We address this issue by introducing the Correlation Aware Pruner (CAP), a new unstructured pruning framework which significantly pushes the compressibility limits for state-of-the-art architectures. Our method is based on two technical advancements: a new theoretically-justified pruner, which can handle complex weight correlations accurately and efficiently during the pruning process itself, and an efficient finetuning procedure for post-compression recovery. We validate our approach via extensive experiments on several modern vision models such as Vision Transformers (ViT), modern CNNs, and ViT-CNN hybrids, showing for the first time that these can be pruned to high sparsity levels (e.g. $\geq 75$%) with low impact on accuracy ($\leq 1$% relative drop). Our approach is also compatible with structured pruning and quantization, and can lead to practical speedups of 1.5 to 2.4x without accuracy loss. To further showcase CAP's accuracy and scalability, we use it to show for the first time that extremely-accurate large vision models, trained via self-supervised techniques, can also be pruned to moderate sparsities, with negligible accuracy loss.

Objective: The purpose of this study was to investigate alpha power as an objective measure of effortful listening in continuous speech with scalp and ear-EEG. Methods: Scalp and ear-EEG were recorded simultaneously during presentation of a 33-s news clip in the presence of 16-talker babble noise. Four different signal-to-noise ratios (SNRs) were used to manipulate task demand. The effects of changes in SNR were investigated on alpha event-related synchronization (ERS) and desynchronization (ERD). Alpha activity was extracted from scalp EEG using different referencing methods (common average and symmetrical bi-polar) in different regions of the brain (parietal and temporal) and ear-EEG. Results: Alpha ERS decreased with decreasing SNR (i.e., increasing task demand) in both scalp and ear-EEG. Alpha ERS was also positively correlated to behavioural performance which was based on the questions regarding the contents of the speech. Conclusion: Alpha ERS/ERD is better suited to track performance of a continuous speech than listening effort. Significance: EEG alpha power in continuous speech may indicate of how well the speech was perceived and it can be measured with both scalp and Ear-EEG.

Cilj. Cilj ovog teorijskog istraživanja jest kritički prikazati suvremene oblike pisme- nosti 21. stoljeća, specifičnosti po kojima su slični ili se razlikuju, s posebnim fokusom na medijsku pismenost kao element medijskog obrazovanja kako bi se pridonijelo stvaranju pretpostavki za oblikovanje koncepta kritičke autonomije. Komplementarnost i sinergija tih pojmova prepoznati su kao ključni čimbenici za uspješan razvoj i obrazovanje svakog pojedinca. Metodologija. U istraživanju se polazi od pretpostavke da je unaprjeđenje suvremenih oblika pismenosti ključno za razvoj društva te zahtijeva od pojedinca sustavno i organizirano stjecanje raznovrsnih kompetencija s ciljem unaprjeđenja veće razine zna- nja i obrazovanja. Stoga se istraživačko pitanje odnosi na koherentnost tih pojmova. Suvremeni oblici pismenosti trebali bi se tretirati kao srodni, komplementarni pojmovi (nikako konkurentni/suparnički), koji se sjedinjuju i nadopunjuju jedni na druge čineći tako sinergiju i zajedničko djelovanje koje za cilj ima prikupljanje novih znanja i stjecanje kompetencija u suvremenom svijetu medija, informacija i obrazovanja. Postdigitalni izazov opisuje odnos pojedinca i tehnologije, ali (bi) treba(o) uključiti i odnos etike, estetike i emocije. Za pripremu i teorijsku analizu u radu korištena je relevantna znanstvena i stručna literaturna građa, s ciljem tumačenja aktualnih suvremenih oblika pismenosti, s posebnim osvrtom na medijsku pismenost u 21. stoljeću i ukazivanja na usklađenost sadržaja i teme istraživanja. Rezultati. Rad donosi kritičku analizu suvremenih oblika pismenosti i to, u metodološkom smislu, pozicioniranih u kontekstu UNESCO-vih definicija pismenosti kao vještina koje u 21. stoljeću predstavljaju uvjet za kvalitetnije obrazovanje. Sukladno definiranim ciljevima, ovaj rad opisuje aktualne suvremene oblike pismenosti prikazujući ih kroz elementarnu, sekundarnu i tercijarnu pismenost. Predstavljena su različita tumačenja suvremenih oblika pismenosti, ističući njihove specifičnosti, s posebnim fokusom na UNESCO-vu strategiju za pismenost mladih i odraslih. Posljednji dio rada fokusiran je na ulogu medijske pismenosti kao elementa medijskog obrazovanja s ciljem da se pridonese kreiranju koncepta kritičke autonomije. U radu se zaključuje da, uz ostale suvremene oblike pismenosti, medijsko opismenjavanje nudi najbolje rezultate ako se odvija u sinergiji s drugim suvremenim i komplementarnim pismenostima 21. stoljeća. Ograničenja. Informacijsko okruženje sa sobom nosi značajne mogućnosti otkri- vanja novih kreativnih potencijala, najčešće u kontekstu razvijanja tehnologije, sociokulturnog prilagođavanja pojedinca društvu i unaprjeđenja društvenog razvoja. U isto vrijeme mnogi brzomijenjajući obrasci društvenog života, prije svega tehnološke inovacije i tržišna konkurentnost, predstavljaju izazove za društvo i pojedinca. Koncept suvremenih pismenosti, s posebnim fokusom na medijsku pismenost, prema tome, predstavlja dinamičnu i promjenjivu komponentu. Medijska pismenost proučava se kao dio procesa cjeloživotnog učenja (odnosi se na svaku aktivnost učenja tijekom cijelog života) te su vještine i kompetencije koje pojedinac stječe podložne promjenama    i uvjetovane su tehnološkim inovacijama koje uključuju nove pristupe informacijama, prijenos informacija, poruka i znanja. Iz tog razloga obrazovanje za suvremene oblike pismenosti predstavlja stalan i promjenjiv proces koji zahtijeva kontinuiranu kritičku evaluaciju tehnologije i medija, kontinuirano stjecanje novih vještina i kompetencija, ali i stvara okruženje za kreiranje novih kreativnih obrazovnih i odgojnih strategija za razumijevanje i korištenje suvremenih tehnologija i medija. Društvena vrijednost. Rad je usmjeren na kritičku analizu suvremenih oblika pismenosti i to, u metodološkom smislu, pozicioniranih u kontekstu UNESCO-ovih de- finicija pismenosti kao vještina koje u 21. stoljeću predstavljaju uvjet za opstanak i kvalitetniju čovjekovu egzistenciju. Poseban osvrt je na medijskoj pismenosti jer ta komponenta, onima koji je posjeduju, omogućuje uspješniju zaštitu njihovih zakonskih prava, sloboda i razumijevanja funkcije medija. Ako je funkcija medija i način na koji mediji stvaraju utjecaj na ljude jasnija, time se konstantno stječe veća kontrola nad tim procesom i ne dopušta se medijima da od svojih korisnika stvaraju pasivne medijske recipijente. Pri tome se ne misli da je pojedinac medijski pismen ako je sposoban primati medijske poruke jer to nužno ne znači da ih je u stanju kritički procjenjivati. Originalnost/vrijednost. Znanstveni doprinos utemeljen je na istraživanju i analizi relevantne znanstvene i stručne literature s ciljem tumačenja aktualnih suvremenih oblika pismenosti s posebnim osvrtom na medijsku pismenost. S obzirom na provedeno istraživanje i analizu, predstavljeni su relevantni čimbenici koji upućuju na to da je sinergija i komplementarnost suvremenih oblika pismenosti ključan element za unaprjeđenje obrazovanja svakog pojedinca.

Airports negatively affect the environment by generating annoying aircraft noise. In practice, negative noise effect due to constant increase of air traffic is being compensated for through improved approach and departure procedures as well as new methods of noise reduction, mainly at the noise source itself. The emergence of COVID-19 pandemic significantly reduced air traffic throughout the world, which had a positive effect on the noise exposure of the people living around airports. This research is analyzing the effect of the COVID-19 on the noise exposure of the population living around Split Airport. The analysis was conducted based on the comparison of the noise maps for the peak days of 2019 and 2020, created in the Integrated Noise Model. The estimated number of people highly annoyed during the day for the peak day of 2020 was reduced by 14.7% compared to 2019, while the estimated number of people highly sleep-disturbed decreased by 62.3%.

BACKGROUND: Tumors of the central nervous system comprise a wide range of over 100 histological distinct subtypes with different descriptive epidemiology, clinical features, treatments, and outcomes. The presence of isocitrate dehydrogenase gene mutation 1 (IDH1) has become one of the most critical biomarkers for molecular classification and prognosis in adult diffuse gliomas. About 65–90% of patients with adult diffuse gliomas have seizures as their initial symptoms. AIM: The objective of this study was to determine the association between IDH1 mutations in adult diffuse gliomas with an incidence of symptomatic epilepsy. METHODS: The study was conducted as an observational, cross-sectional, and prospective clinically controlled study at the Clinic of Neurosurgery of the Clinical Center of the University of Sarajevo. The research included a total of 100 patients treated at the Clinic of Neurosurgery, with pathohistological confirmation of glioma Grades II–IV who were stratified by groups according to tumor grade. Data were collected on tumor localization and grade, the presence of IDH mutations, and the presence of epileptic seizures as the first symptom of the glioma. RESULTS: Out of a total of 100 patients, 39 had IDH 1 mutations, while 61 patients were without them: Of these, diffuse astrocytoma Grade II 30 cases (30%), Grade III 5 (5%), and Grade IV 7 (7%), and the number of patients with glioblastoma was 58 (58%). In the group of patients with IDH 1 mutations, epileptic seizures were present in 87.2% compared to the group of patients without IDH 1 mutations (wild type) in which epileptic seizures were present in 16.4% of cases. Statistical analysis showed that the positive mutated IDH-type carries an almost 70% increase in the likelihood of epileptic seizures (χ2 = 8.378; p = 0.0001). If we separate the group of diffuse astrocytomas in the IDH 1-positive subgroup, 34 patients (85.81%) had epileptic seizures, while in the IDH 1-negative subgroup, there were no patients with epileptic seizures, which carries a statistically significant difference in frequency in favor of IDH 1-positive tumors (p ≤ 0.001). CONCLUSION: There is a clear connection between the presence of IDH1 mutations and the occurrence of epileptic seizures in the clinical picture of patients with diffuse adult glioma.

We present stability conditions for deterministic time-varying nonlinear discrete-time systems whose inputs aim to minimize an infinite-horizon time-dependent cost. Global asymptotic and exponential stability properties for general attractors are established. This work covers and generalizes the related results on discounted optimal control problems to more general systems and cost functions.

Strategies to mitigate urban heat islands are a recent issue in the Austrian capital, Vienna. In this study, the uhiSolver-v2106-0.21 software was used to evaluate the summer cooling effects and humidity production of small-scale facade greening and a green pergola located in two schools within the city. Based on on-site measurement data, the study revealed that small-scale greening measures are not able to substantially reduce ambient air temperature. On a hot summer day, at 3 p.m. local time (CEST), the maximum decrease amounted to 0.3 °C at 0.1 m from the facade greening as well as inside the green pergola. As for the apparent (perceived) temperature, a reduction of up to 4 °C was observed under the green pergola compared to the unshaded roof terrace. Hence, the simulation results show that, within urban areas, a significant improvement of thermal comfort in summer can only be achieved through large-scale greenery that provides shade for pedestrians.

The subject of the research is the analysis of services that should be offered by public administration toward legal entities and natural persons. We will suggest that it uses real-time intelligent IT systems based on data storage based on the data warehouse. Our opinion is that such systems implemented in the public administration of Bosnia and Herzegovina could offer an appropriate service to each user at a time when they need the service in a form that suits them, on the spot where he is in the moment of delivery and in the quantity that he needs. In order for a public administration to be able to offer an appropriate service, it must know its users much better and know which service they need at what time. In this paper, we analyze self-service as well as the necessary activities for all that. These are all activities that should be carried out in the future, and this work could contribute to the implementation of these processes.

We revisit the performance of the classic gradual magnitude pruning (GMP) baseline for large language models, focusing on the classic BERT benchmark on various popular tasks. Despite existing evidence in the literature that GMP performs poorly, we show that a simple and general variant, which we call GMP*, can match and sometimes outperform more complex state-of-the-art methods. Our results provide a simple yet strong baseline for future work, highlight the importance of parameter tuning for baselines, and even improve the performance of the state-of-the-art second-order pruning method in this setting.

The individual interpretations of purchasing policies weaken fair government purchasing practices. However, this does not fully account for the nature of the media’s coverage of government contracting, taken as a whole. The authors seek to understand better framing in newspaper stories on government procurement, with the government as a force that creates opportunity and fraud. The paper focuses on an area of cognitive uncertainty in understanding portrayals of public procurement in newspaper articles as positive or negative, and, assuming a lack of balance, what interest or group do articles favor in their portrayal of this public function? Sentiment analysis of a corpus of newspaper articles focusing on government contracting was conducted. This analysis suggests that the negative perception of government contracting is reinforced and exacerbated by sensationalized media coverage, a negative impact on the policymaking process and public discourse, and public trust in government results.

Endurance-type disciplines (running, cycling, biathlon) define the cyclic structure of an athlete's movements, which, in addition to functional parameters, also includes an adequate morphological profile and body composition. Based on the detection, analysis and evaluation of these parameters, it is possible to define the body composition of the competitors as well as possible mutual differences even though it is endurance sports. The results are all the more relevant if the profile of top athletes with notable results is being evaluated. The current case study analyzes the morphological dimensions and body composition of competitors of three different disciplines (middle and long distances, cycling, biathlon) of top-level competitors, members of national teams. The study was conducted: Uroš Gutić (UG) - runner middle and long distances, member of AK "Sarajevo" and the BIH athletic national team; Milan Milivojević (MM) – cyclist, member of Cycling club "Borac" Čačak (Serbia), and the member Serbian national team; Stefan Lopatić (SL) – biathlete, member SK "Romanija" Pale, and BIH national team.

The paper proposes an objective image quality assessment measure with full referencing. The measure is based on a comparison of the contrast of the original image and the image with the degradation. Discrete cosine transform coefficients are used for contrast estimation. By applying the measure, a scalar value is obtained that reflects the quality of the test (degraded) image. The proposed measure is tested on an infrared image dataset developed by the Military Academy in Belgrade, Serbia. The performance of the measure was compared with other well-known objective full-reference image quality assessment metrics, which were developed for the images in visible domain. It was shown that measure performance can be improved with the adequate selections of the block dimensions and the number of discrete cosine transform coefficients during the calculation of image quality value. The proposed measure obtained a correlation with the subjective scores near 82%, which puts the measure into the top three of all tested image quality assessment measures. The proposed measure showed the best performance on the images distorted by Gaussian blurring, where the level of agreement with the subjective scores is over 97%, according to which the measure stands out as the best compared to other tested measures.