Croatian Digital Dissertations Repository
Not a member yet
10487 research outputs found
Sort by
Arhitektura sustava za automatsko raspoređivanje poslova na heterogenim računalima visokih performanci zasnovana na metodama strojnoga učenja i dinamičkim parametrima rada
No full textOne of the most complex problems in high-performance computing (HPC) is the efficient allocation and management of resources. Performance, ease of use, and energy efficiency are key design goals. These goals are often opposed, making optimization challenging within the same system.
This dissertation explores the problem of optimizing HPC systems. It proposes a new machine learning approach to dynamically evaluate workload parameters and automatically schedule jobs within an HPC environment based on the Kubernetes platform.
Existing methods, including the use of virtual machines and physical servers, were analyzed, and shortcomings such as scalability, energy efficiency, and flexibility were identified. Containerization (Docker/Podman) was highlighted as a modern solution which brings challenges in resource management, network latency, and energy consumption optimization.
The proposed system includes algorithms for dynamic estimation of the availability of computing nodes and optimized workload allocation using machine learning. The implementation of the system is based on the Kubernetes platform, which can work in heterogeneous environments (x86, ARM, RISC-V, NVIDIA GPU). Automation of deployment and configuration is enabled through Ansible, while the system's modularity allows for flexible integration of different network and resource configurations.
The developed software platform allows users to efficiently run HPC applications with full automation or manual management as needed. The use of custom algorithms and the integration of an HPC workload manager with Kubernetes have improved operational efficiency, enabling more optimal resource use in dynamic HPC systems.Jedan od najsloženijih problema u računarstvu visokih performansi (HPC) je efikasno raspoređivanje i upravljanje resursima, pri čemu su ključni ciljevi dizajna performanse, jednostavnost korištenja i energetska učinkovitost. Ovi ciljevi često suprotstavljeni, čineći optimizaciju unutar istog sustava izazovnom.
U ovoj se disertaciji istražuje problem optimizacije HPC sustava i predlaže novi pristup korištenjem strojnog učenja za dinamičku procjenu parametara radnog opterećenja i automatsko raspoređivanje poslova unutar HPC okruženja temeljenog na Kubernetes platformi.
Analizirane su postojeće metode, uključujući korištenje virtualnih mašina i fizičkih poslužitelja, te su identificirani njihovi nedostaci poput skalabilnosti, energetske učinkovitosti i fleksibilnosti. Kao suvremeno rješenje istaknuta je kontejnerizacija (Docker/Podman), koja donosi izazove u upravljanju resursima, mrežnoj latenciji i optimizaciji energetske potrošnje.
Predloženi sustav uključuje algoritme za dinamičku procjenu dostupnosti računalnih čvorova i optimizirano raspoređivanje radnih opterećenja koristeći strojno učenje. Implementacija sustava temelji se na Kubernetes platformi, s mogućnošću rada u heterogenim okruženjima (x86, ARM, RISC-V, NVIDIA GPU). Automatizacija implementacije i konfiguracije omogućena je putem Ansible-a, dok modularnost sustava omogućuje fleksibilnu integraciju različitih mrežnih i resursnih konfiguracija.
Razvijena programska platforma omogućuje korisnicima jednostavno pokretanje HPC aplikacija uz potpunu automatizaciju ili ručno upravljanje prema potrebama. Korištenjem prilagođenih algoritama i integracijom HPC upravitelja radnog opterećenja s Kubernetesom poboljšana je operativna učinkovitost, čime se omogućuje optimalnije korištenje resursa u dinamičnim HPC sustavima
New features of interviews in prime time television news in Croatia during the coronavirus pandemic
No full textNovinarski intervju jedan je od najvažnijih oblika novinarskog izražavanja, ali i oblika javne komunikacije izabranih sugovornika. U novinarstvu pitanja se postavljaju s točno određenim ciljem i svrhom kako bi se otkrilo korisne i zanimljive činjenice, dobilo ideje, tvrdnje, gledišta, ali i privuklo ili predstavilo važne, relevantne i zanimljive sugovornike, također, dobilo i mozaik informacija čime, bilo da je intervju samostalan oblik ili forma, bilo da upotpunjuje ostale novinarske oblike, obogaćuje medijski sadržaj.
Cilj je ove doktorske disertacije istražiti promjene u učestalosti, formi i sadržaju televizijskog intervjua u 2020., pandemijskoj godini u odnosu na 2019., neposredno prije pandemije, u dnevnim središnjim informativnim emisijama triju nacionalnih televizija. Prva pandemijska godina, naime, značajno je utjecala na dotadašnje redakcijske prakse, zbog djelomičnog i potpunog zatvaranja zemlje te epidemioloških pravila, što je i glavni razlog promjena u strukturi i novinarskim oblicima od kojih se dnevne središnje informativne emisije sastoje.
Istraživanje je provedeno metodom kvantitativne i kvalitativne analize sadržaja televizijskih informativnih emisija na dvije komercijalne televizije (RTL i Nova TV) i na javnoj televiziji (HTV) pri čemu se analiziralo jesu li i kako izvanredne okolnosti, kao što je pandemija koronavirusa u Hrvatskoj, utjecale na frekventnost i obilježja dotadašnjeg izvještavanja u dnevnim središnjim informativnim emisijama.
Ukupno je analizirano 210 emisija od siječnja do prosinca 2019. i 2020. Rezultati analize pokazali su da je u 2020. glavna tema pandemija, ali da su i ostale dominantne teme obrađene u kontekstu pandemije za razliku od 2019. kada je dominantna tema bila unutarnja politika, i to u istom postotku kao i COVID-19 u pandemijskoj godini.
Istraživanje je pokazalo i da su najčešći oblici izvještavanja bili izvještaji i intervjui pri čemu je u pandemijskoj godini bilo manje izvještaja u odnosu na 2019., a dijelom je povećan broj intervjua u odnosu na godinu prije, ponajprije zbog lockdowna. Zanimljivo je pritom, kako pokazuje istraživanje, da je u 2020. učestaliji i hibridni oblik televizijskog izvještavanja – javljanje uživo radi intervjua u odnosu na godinu prije, jer je to bio jedan od najefikasnijih načina da se u novim okolnostima dođe do novih informacija i dodatnih objašnjenja događaja i tema.The journalistic interview is a vital form of journalistic expression and a key medium for public communication with selected interlocutors. In journalism, questions are posed with specific goals and purposes: to uncover useful and interesting facts, generate ideas, present viewpoints, and attract or highlight significant and engaging interlocutors. Whether as a standalone format or complementing other journalistic forms, interviews enrich media content by providing a mosaic of information.
This doctoral dissertation examines changes in the frequency, format, and content of television interviews during the pandemic year of 2020 compared to 2019, the year preceding the pandemic. The focus is on the central daily news programs of three national television networks. The pandemic's first year significantly altered editorial practices due to partial and full lockdowns and epidemiological restrictions, leading to notable changes in the structure and forms of journalism featured in central news programs.
The research employs quantitative and qualitative content analysis of central news programs on two commercial networks (RTL and Nova TV) and the public broadcaster (HTV). It investigates whether and how extraordinary circumstances, such as the COVID-19 pandemic in Croatia, affected the frequency and characteristics of reporting in these news programs. A total of 210 programs, spanning January to December in 2019 and 2020, were analyzed. The results show that the dominant topic in 2020 was the pandemic, while other significant topics were often framed within its context. In contrast, the dominant topic in 2019 was domestic politics, which accounted for the same percentage of coverage as COVID-19 in the pandemic year.
The study also found that the most common reporting formats were news reports and interviews. In 2020, the number of news reports decreased compared to 2019, while the frequency of interviews slightly increased, primarily due to lockdown conditions. Interestingly, as the research indicates, 2020 saw a notable increase in the use of hybrid television reporting formats—specifically live reporting for interviews—compared to the previous year. This approach emerged as one of the most effective ways to gather new information and provide additional explanations of events and topics under the new circumstances. In Croatia's pandemic experience, news content was produced within newsrooms operating under altered conditions and strict epidemiological measures. Newsroom workflows were reorganized, with teams working under reduced resources, making it impossible to produce reports in their previous scope and format. During the lockdown, statements from contributors for news reports were recorded using platforms such as Zoom, Microsoft Teams, and Skype. Statements from citizens, meanwhile, were recorded from balconies, windows, or other locations that ensured sufficient physical distance.
Changes in news production during the pandemic year were reflected in shifts in television journalistic formats, as revealed by the quantitative content analysis. Overall, significant changes in the frequency of interviews were identified, along with several specific characteristics:
1) Topics – Given that 2020 was predominantly a pandemic year, the main topic of coverage was unsurprisingly COVID-19. In contrast, the primary focus in 2019 was domestic national politics, accounting for as much as 40% of the total coverage.
2) Increased representation of interviews and live reporting in headlines – the results show that during the pandemic year, television networks more frequently and clearly announced interviews, interlocutors, and the journalists conducting them in live reports compared to 2019. This reflects the heightened importance placed on these formats in 2020.
3) New interlocutors – the pandemic introduced an entirely new group of interlocutors and communicators, including numerous previously unknown figures such as doctors, epidemiologists, immunologists, seismologists, and scientists. Additionally, representatives from various walks of life—hospitality workers, small business owners, farmers, entrepreneurs, artists, and athletes—were featured, as the pandemic disrupted or significantly hindered their daily work and lives.
4) Interview locations during the pandemic due to the restricted functioning of newsrooms – the majority of interviews were conducted in the field, at outdoor locations: more than 60% in 2020 compared to 40% in 2019. Studio interviews were most frequently held on Dnevnik 2 on HTV (Croatian National Television), which had the most suitable conditions, primarily due to the size of the studio and its separation from the newsroom.
We specifically observed particular features in live reporting. Notably, journalists more frequently reported live exclusively for interviews, which emerged as a crucial journalistic format during this period. Live reporting was also more justified in 2020 compared to 2019. An interesting trend was the monthly fluctuation in live reporting for interviews during the pandemic year. The percentage increased during lockdowns, when access to television studios was restricted, dropped in the summer as epidemiological measures eased, and rose again in the fall and winter as infection rates climbed.
The study also identified differences between commercial networks (Nova TV and RTL) and the public broadcaster (HTV). During the peak months of the pandemic, Nova TV and RTL conducted more live interviews overall, especially as part of live reporting, compared to HTV. Nova TV had the highest number of live reports for interviews, nearly 65%. In the comparative year, 2019, Dnevnik 2 and RTL Danas were dominated by live reporting compared to interviews and live reports for interviews. In contrast, Dnevnik Nova TV had nearly 62% of its live reports dedicated to interviews, while HTV had 27% and RTL 37%.
The research confirmed three hypotheses: H1: The pandemic increased the frequency of interviews in central news programs. H2: Extraordinary circumstances statistically increased the use of hybrid formats, particularly live reporting for interviews. H3: National television networks emphasized interviews in their headlines more during the pandemic than before.
The news programs of all three television networks largely reported on current and engaging topics during the pandemic through interviews, either in studio formats or as part of live reporting. In addition to traditional live reporting, characterized by journalists delivering new information from the scene of an ongoing or recent event, a hybrid journalistic format—live reporting for interviews—became more frequent and significant, primarily due to epidemiological measures. During lockdowns, when interlocutors were unable to visit television studios, many interviews were conducted at outdoor locations, enabled by technological solutions. In live reporting, the primary role of the journalist appearing on camera is to convey unfolding events or those that have just occurred, provide the latest information about a specific event, analyze, and capture the atmosphere (Perišin, 2010). The inclusion of interlocutors in live reporting added significant value, especially in the context of the crisis. Technology and the development of the internet today allow for live broadcasts from any location with a stable internet connection, which proved indispensable under pandemic conditions. Overall, this study highlights the irreplaceable role of television in covering large scale events. Its fundamental value—"live television"—was especially evident during the pandemic through daily live broadcasts of press conferences by the National Security Headquarters and the timely dissemination of information via interviews and live reports in news programs.
Journalists conducting live reports for interviews served as crucial intermediaries between the television studio (and the news program host) and the interviewee. Typically, the journalist does not introduce new information or atmosphere themselves but instead asks or reiterates questions from the host and follows up with additional questions. Sometimes, these interviews do not yield new information; rather, they serve to fill airtime or demonstrate the network's technical capabilities (Cushion & Lewis, 2010; Perišin, 2010). However, during the pandemic year, such live reporting was justified, as interlocutors became an essential part of the content in central news programs. While the qualitative analysis of six selected interviews during the pandemic year showed that all achieved their objectives, providing new information and perspectives, there remains room for improvement in the substance and conceptualization of interviews.
All the selected interviews fulfilled their purpose and objectives, providing new information and perspectives through the conversations. None of the interviews were conducted merely to fill airtime. Journalists demonstrated significant skill in asking the right questions and showed an understanding of the strategy behind the first question, which is crucial for the flow and success of the entire interview.
However, one notable shortcoming in television interviewing was the introduction of interviewees, including their names and titles. In many of the interviews analyzed quantitatively, journalists failed to mention who they were speaking with, instead deferring to phrases like "our guest will address this" or referring to the individual by first name, last name, or general terms like "Mr." or "Ms." without further context. This issue was also apparent in two interviews examined qualitatively, where the interviewees were only addressed as "Ms." or "Mr." followed by their first name, with no additional details provided.
Another area for improvement is the use of language and speech. Common issues include reliance on foreign words, colloquialisms, and jargon. While many questions were substantively on track, they often lacked proper grammatical structure in Croatian.
Another area for improvement is the use of language and speech. Common issues include reliance on foreign words, colloquialisms, and jargon. While many questions were substantively on track, they often lacked proper grammatical structure in Croatian.
The most important aspect for journalists in Croatia to focus on in television interviews is constructiveness. Even when events carry negative informational factors, these do not need to dominate the narrative from beginning to end. This is echoed by the audience, who expect higher-quality content from the media, as shown by the public perception survey conducted as part of the Jourlab project (2021). Constructive journalism provides a partial answer to audience expectations. Tena Perišin and Nikola Šimić (2020) found that when events contain elements of negativity, these do not need to dominate the narrative from beginning to end, and interviews can play a crucial role in addressing this. Their research in Scandinavian countries revealed that journalists advocating for constructive journalism consider fairer and more thoughtful interview questions a key step forward in enhancing the quality of journalism as a whole (Perišin & Šimić, 2020, p. 207).
In conclusion, this study introduces a novel methodological approach to examining the concept of interviews within central television news programs—a topic previously unexplored in Croatia. The scientific contribution of this work lies in the use of both quantitative and qualitative methods of analysis for the first time, providing a more comprehensive understanding of the significance of television interviews as a segment of news programs. The research findings will enhance scientific understanding of the characteristics of this journalistic genre and the role of interviews in news programs. Additionally, they address the question of how interviews, as part of a broadcast, contribute to the creation of media content. Along with its academic value, the results will also have practical applications in journalistic practice
Sinteza i implementacija pasivnih zakrivljenih metapovršina
No full textMetasurfaces have garnered significant research interest since their emergence at the turn of the century. Proposed applications aim to implement electromagnetic functions in a compact form. Many applications will require curved structures that conform to the application environment.
This thesis aims to provide a concise framework for analysing, designing, and producing passive curved metasurfaces. Several analysis methods are proposed to determine the response of curved metasurfaces. A new Resonance Cancellation Method was introduced, which simplifies the analytical approach for general curvatures. For metasurface design, we review several unit cell designs and verify the expressions for surface impedances. Given the production limitations, as well as practical constraints observed in designing, we provide guidelines on selecting unit cell designs. Along with producing prototypes in printed circuit board technology, we explored alternative production methods, testing metallic sprays to create functional metasurfaces.
We designed several cylindrical metasurface prototypes, used to modify and shape the beam of an existing antenna system enveloped by the metasurface. The expected responses were analysed using the methods outlined in this thesis, and designs followed the established guidelines. Performance of manufactured metasurfaces was measured to confirm alignment with theoretical predictions.Metapovršine privlače značajan istraživački interes od svog prvog pojavljivanja na prijelazu stoljeća. Predloženim primjenama žele se implementirati određene elektromagnetske funkcije u kompaktnom obliku. Mnoge će primjene zahtijevati zakrivljene strukture oblikovane prema specifičnim zahtjevima.
Ovom disertacijom želi se pružiti koncizan okvir za analizu, dizajn i proizvodnju pasivnih zakrivljenih metapovršina. Predlaže se nekoliko metoda za analizu kojom se određuje odziv zakrivljenih metapovršina. Uvodi se nova metoda poništavanja rezonancija koja pojednostavljuje analitički pristup za općenito zakrivljene strukture. Za dizajn metapovršina, daje se pregled nekoliko dizajna jediničnih ćelija i verificiraju se izrazi za površinsku impedanciju. S obzirom na ograničenja proizvodnih tehnologija, kao i praktična ograničenja primijećena tijekom dizajniranja, daju se smjernice za odabir dizajna jediničnih ćelija. Uz izradu prototipa u tehnologiji tiskanih pločica, istražene su i alternativne metode proizvodnje, ispitujući metalizirajuće sprejeve za izradu funkcionalnih metapovršina.
Dizajnirano je nekoliko prototipa cilindričnih metapovršina namijenjenih oblikovanju dijagrama zračenja postojećeg antenskog sustava okruženog metapovršinom. Očekivani odzivi analizirani su metodama opisanima u ovoj disertaciji, a dizajni su slijedili utvrđene smjernice. Performanse izrađenih metapovršina izmjerene su kako bi se potvrdilo njihovo poklapanje s teorijskim predviđanjima
Razvoj neutronskih senzora za nalaženje sirove nafte, metan hidrata i elemenata rijetkih zemalja
No full textThis dissertation explores the development and application of neutron-based sensors for detecting crude oil, methane hydrates, and rare earth elements (REEs), employing both experimental measurements and Monte Carlo (MC) simulations. The study aims to im prove and explore elemental detection techniques by utilizing the associated alpha particle (AAP) method and pulsed fast-thermal neutron activation (PFTNA), finding detection limits and optimizing background suppression. In the field of crude oil exploration, a compact neutron probe was designed and tested for carbon-to-oxygen (C/O) logging using the AAP method. Experimental results demon strated that AAP significantly reduces background noise compared to conventional fast neutron activation analysis (FNAA). The probe’s performance was tested under high temperature conditions, while its temporal and spatial resolutions were determined to be 2 ns and 10 cm, respectively. The probe was further utilized in PFTNA mode to determine the minimum detection limits (MDLs) for chlorine and water content in crude oil, yielding (62±6) mg/L and (2±2) vol.%, respectively. While the chlorine MDL was within the required export limits of 50 mg/L, the water MDL was inconclusive due to high relative uncertainties. Methane hydrate detection was conducted using the AAP technique in an experimental setup designed to simulate oceanic hydrate formations. Key components, such as compact neutron generator, detectors and powers supplies were mounted in the submarine. The MDL for methane hydrate saturation was determined to be (28±8) vol.%, which is suitable for detecting high-saturation hydrate reservoirs such as those found near Mount Elbert in Alaska. The final section of this work focuses on the detection of REEs, specifically gadolin ium and lutetium, which serve as indicators of total REE concentrations in deep-sea sediments. For gadolinium, an active neutron-based approach leveraging its high neutron capture cross-section was used, achieving an MDL of (12±1) ppm in the gamma-energy region (2.55–3.3 MeV), sufficient for detecting concentrations in both the Pacific (68±2 ppm) and Indian (39±2 ppm) oceans. Lutetium detection relied on passive gamma-ray spectroscopy with optimized lead shielding, achieving an MDL of (3.8±0.7) ppm for a 50-hour measurement, which is sufficient for Pacific Ocean sediments but would require longer integration times for Indian Ocean samples. Monte Carlo simulations were employed across all applications to validate experimen tal results, optimize detection configurations, and investigate the underlying factors con tributing to certain experimentally observed phenomena. The methodologies developed in this dissertation demonstrate the feasibility of neutron-based techniques for remote and in-situ elemental detection. Future work should focus on further optimizing neutron colli mation, improving detector resolution, and integrating neutron sensors into autonomous underwater vehicles (AUVs) or remotely operated vehicles (ROVs) for real-world applica tions in deep-sea exploration.Metode neutronske analize omogućuju precizno određivanje elementarnog sastava različitih materijala, čime imaju široku primjenu u industriji, geologiji i fundamentalnoj fizici. Ova disertacija istražuje primjenu triju glavnih neutronskih metoda – Fast Neutron Activation Analysis (FNAA), Associated Alpha Particle (AAP) metode i Pulsed Fast-Thermal Neutron Activation (PFTNA) – u svrhu detekcije sirove nafte, metan hidrata i elemenata rijetkih zemalja. Eksperimentalni postavi razvijeni su za specifične scenarije karakterizacije uzoraka korištenjem neutronskog zračenja, dok su Monte Carlo simulacije provedene kako bi se optimizirala geometrija eksperimenta i bolje razumjeli eksperimentalni rezultati. Ova istraživanja predstavljaju doprinos razvoju tehnika zasnovanih na neutronima za daljinsku i in-situ karakterizaciju materijala u industrijskim i prirodnim okruženjima
Međudjelovanje električno pobudljivih stanica i fotokondenzatora
No full textThis thesis investigates the interaction between electrically excitable cells and organic electrolytic photocapacitor (OEPC) devices, with the goal of enhancing photocapacitive stimulation for neurostimulation applications. The study takes a multidisciplinary approach, combining materials science, biophysics, and computational modeling to refine OEPC design and better understand the mechanisms underlying their interaction with biological tissues. The original research contributions include improvements in OEPC device design and geometry, optimization of organic semiconductor layer thickness, and the incorporation of advanced return electrode materials such as PEDOT:PSS. This thesis presents a key discovery: 2D structuring and the adoption of an "inverted" OEPC geometry can enhance electric field strength by up to 20 times, enabling precise single-cell stimulation with high spatial resolution. A comprehensive numerical model was developed to simulate extracellular stimulation of excitable cells, incorporating three subunits: the device model, the electrode-electrolyte interface model, and the target cell model. Finite element simulations demonstrate that capacitive microelectrodes exhibit lower stimulation thresholds than Faradaic electrodes when in close proximity to the target cell. The model also validates that photocapacitive devices can effectively stimulate neurons under sufficiently high illumination intensities, provided the cells are tightly attached to the device surface. This research concludes that OEPCs, especially those with inverted geometries and PEDOT:PSS coatings, significantly enhance spatial resolution and stimulation efficiency, highlighting their potential for wireless, minimally invasive neurostimulation applications.Cilj ovog doktorskog rada je detaljno istražiti međudjelovanje fotokondenzatorskih uređaja i električki pobudljivih stanice poput neurona. Ovoj temi pristupamo korištenjem različitih pristupa, od optimizacije samog uređaja u vidu dizajna, geometrije i odabira materijala, biofizičkim modeliranjem stanica radi boljeg razumijevanja praga aktivacije potrebnog za učinkovitu stimulaciju, modeliranjem sučelja između elektrode i stanice kako bi se osiguralo djelotvorno prenošenje podražaja te u konačnici i izradom samih uređaja. Kako bi se svi ovi aspekti mogli u potpunosti istražiti, nužno je usvojiti multidisciplinarni pristup koji uključuje stručnost iz područja biofizike, znanosti o materijalima i medicin
EVALUATION OF EFFECTS OF SMOKELESS TOBACCO AEROSOL ON ORAL CAVITY STRUCTURE AND COMPARISON WITH CONVENTIONAL TOBACCO PRODUCTS
No full textCiljevi: Procijeniti i usporediti morfološki izgled oralne sluznice, prevalenciju kandidijaze,
prisutnost neugodnog zadaha, olfaktornu i gustatornu funkciju, specifične karakteristike sline
(količina, biomarkeri upale i oksidativnog stresa - IL-1β, IL-6, IL-2, TNF-α, 8-OHdG) i
psihološke simptome depresivnosti, stresa i anksioznosti u skupinama pušača novih duhanskih
preparata bez izgaranja (Tobacco Heating System, THS), pušača konvencionalnih cigareta i
nepušača.
Ispitanici i metode: U istraživanje je uključeno 30 pušača THS-a, 30 pušača konvencionalnih
cigareta i 30 nepušača. Ispunili su anamnestički upitnik, upitnik o pušenju i psihološke upitnike
za procjenu stresa, depresivnosti i anksioznosti. Napravljeni su klinički oralni pregled i testovi
(test lučenja sline, obrisak oralne sluznice, organoleptička procjena halitoze, testiranje okusne
i mirisne funkcije). Sakupljeni uzorci sline su korišteni za imunoenzimsku analizu
koncentracija salivarnih biomarkera upale i biomarkera oksidativnog stresa.
Rezultati: U obje skupine pušača zabilježena je viša prevalencija morfoloških promjena
sluznice (P < 0,001), smanjena količina nestimulirane sline (P < 0,001) i prisutnost neugodnog
zadaha (P < 0,001) u odnosu na nepušače. Nije pronađena značajna razlika u prevalenciji
kandidijaze. Pušači konvencionalnih cigareta pokazali su smanjenu mirisnu funkciju u
usporedbi s nepušačima (P = 0,001), dok se pušači THS-a nisu razlikovali od pušača
konvencionalnih cigareta niti nepušača. U obje skupine pušača smanjena je sveukupna okusna
funkcija, te identifikacija gorkog, kiselog i slanog okusa (P < 0,001). Koncentracija 8-OHdG-a
je značajno viša u skupini nepušača u odnosu na obje skupine pušača (P < 0,001). Nije
zabilježena razlika u koncentraciji salivarnih citokina. Pušači konvencionalnih cigareta niže
percipiraju stres u usporedbi s nepušačima i pušačima THS-a (P = 0,012). Nije bilo linearne
povezanosti između biomarkera oksidativnog stresa i promatranih psiholoških simptoma.
Zaključak: Pušenje ima negativan učinak na morfološki izgled sluznice, količinu sline,
prisutnost neugodnog zadaha, mirisnu i okusnu funkciju, pri čemu je osjet mirisa značajno više
narušen u skupini pušača konvencionalnih cigareta.Objective: To assess and compare the morphological appearance of the mucosa, prevalence of
candidiasis, presence of halitosis, olfactory and gustatory function, specific saliva attributes
(amount, biomarkers of inflammation and oxidative stress - IL-1β, IL-6, IL-2, TNF-α, 8-OHdG)
and psychological symptoms of depression, stress and anxiety in groups of smokers of new
tobacco products without combustion (Tobacco Heating System, THS), conventional cigarette
smokers and non-smokers.
Patients and methods: 30 THS smokers, 30 conventional cigarette smokers, and 30 non
smokers were included in the study. All subjects completed the anamnestic and smoking
questionnaire, and psychological questionnaires to assess stress, depression and anxiety. Oral
examination and tests were performed (saliva secretion test, swab test, organoleptic assessment
of halitosis, olfactory and gustatory testing). The saliva samples were used for immunoenzyme
analysis of the concentrations of biomarkers of inflammation and oxidative stress.
Results: In both groups of smokers, a higher prevalence of morphological changes (P < 0.001),
a reduction of unstimulated saliva (P < 0.001), and the presence of halitosis (P < 0.001) were
observed. No difference was found in the prevalence of candidiasis. Conventional cigarette
smokers showed reduced olfactory function when compared to non-smokers (P = 0.001), while
THS smokers did not differ from conventional cigarette smokers or non-smokers. Gustatory
function and identification of bitter, sour and salty tastes were reduced in both smokers groups
(P < 0.001). The concentration of 8-OHdG was higher in non-smokers (P < 0.001). No
differences were observed in the concentration of cytokines. Conventional cigarette smokers
reported lower levels of perceived stress compared to non-smokers and THS smokers (P =
0.012). No association was observed between oxidative stress biomarker and psychological
symptoms.
Conclusion: Smoking adversely affects the morphological appearance of the mucosa, saliva
production, the presence of halitosis, olfactory and gustatory function, with the olfactory
function being more impaired in conventional cigarette smokers
Computational model of multimodal information fusion for achieving interaction between human and robot
No full textModeran svijet se sve više oslanja na automatske i robotske sustave. Robotika
postaje sastavni dio života svakog čovjeka. Trenutno ovi sustavi ne mogu interaktirati
s ljudima na razini na kojoj to ljudi rade jedni s drugima. Područje kognitivne
informatike i robotike nastoji unaprijediti tu interakciju. Ljudi za prijenos informacija
i interakciju s drugim jedinkama koriste mnogo metoda, primarno verbalnu
i neverbalnu komunikaciju. Verbalna komunikacija koristi skup zvukova, odnosno
glasova, kako bi prenijela informaciju, dok neverbalna komunikacija koristi suptilne
vizualne, akustične ili lingvističke znakove. Ovaj rad bavi se neverbalnim metodama
interakcije čovjeka i virtualnog agenta, pritom proučavajući metode koje ljudi
koriste kako bi postigli zajedničko razumijevanje i kvalitetnu interakciju. Metode
koje se koriste fokusiraju se na analizu neverbalnih znakova kao što su ekspresije
lica, tonalitet i brzina glasa te suptilne promjene u kontekstu govora. Istražuju se
različiti algoritmi višemodalnog stapanja informacija te interakcijske strategije kako
bi se postigla viša razina interakcije. Skupljeni su setovi podataka koji dodatno
pospješuju rad algoritama. Konačno, predloženi modeli testiraju se u kontroliranim
i nekontroliranim okruženjima te se analiziraju dobiveni rezultati. Ukupno istraživanje
je temeljeno na dvije hipoteze:
• Interakciju između čovjeka i agenta u sklopu stvarne, virtualne te miješane
okoline moguće je temeljiti na sinergijskom djelovanju ograničenog broja osjetilnih
modalnosti stopljenih u jedan računalni model.
• Interakcija čovjeka i agenta u realnom vremenu može se postići koristeći interakcijsku
strategiju koja uzima u obzir trenutne i značajne promijene u sklopu
prostora i vremena.The modern world increasingly relies on automated and robotic systems. Robotics
is becoming an integral part of everyone’s life. Currently, these systems are
unable to interact with humans at the same level as humans interact with each other.
The field of cognitive informatics and robotics aims to enhance this interaction. Humans
use many methods to transfer information and interact with others, primarily
verbal and nonverbal communication. Verbal communication uses a set of sounds,
i.e., voices, to convey information, while nonverbal communication employs subtle
visual, acoustic, or linguistic cues. This study focuses on nonverbal methods of interaction
between humans and virtual agents, examining the methods humans use
to achieve mutual understanding and quality interaction. The methods used focus
on analyzing nonverbal cues such as facial expressions, tone and speed of voice, and
subtle contextual changes in speech. Various algorithms for multimodal information
fusion and interaction strategies are explored to achieve a higher level of interaction.
Datasets have been collected to further enhance the algorithms. Finally, the
proposed models are tested in controlled and uncontrolled environments, and the
obtained results are analyzed. The overall research is based on two hypotheses:
• Interaction between humans and agents within real, virtual, and mixed environments
can be based on the synergistic action of a limited number of sensory
modalities fused into a single computational model.
• Real-time interaction between humans and agents can be achieved using an
interaction strategy that accounts for current and significant changes in space
and time
Svojstva ponašanja primorske gušterice, Podarcis siculus (RafinesqueSchmaltz, 1810)
No full textInvasive species are able to colonise new environments, where they can threaten biodiversity
and have a wide range of negative effects on ecosystems. To understand which behaviours are
associated with competitive success and invasion, I studied behaviour of invasive species
Podarcis siculus in three ecological contexts: 1) on interspecies level with subordinate
Podarcis melisellensis, 2) on intraspecies level between populations inhabiting differing
habitats, and 3) between populations inhabiting similar habitats. Podarcis siculus showed
higher exploratory behaviour, lower boldness, and consumed more food than P. melisellensis.
Podarcis siculus population on a resource-constrained islet exhibited decreased aggressivity
and corresponding phenotypic traits than the mainland population. Populations in similar
habitats showed low differentiation in behaviour. However, substantial behavioural variability
among individuals was observed and certain behavioural traits were correlated. Overall, P.
siculus demonstrated specific behaviours and intra- and interspeciesbehavioural variability
that can be beneficial in both competition and invasiveness.Invazivne vrste sposobne su kolonizirati nova staništa, gdje mogu imati širok raspon
negativnih učinaka na ekosustave te ugroziti bioraznolikost. Kako bi razumjeli koja ponašanja
su povezana s kompetitivnim uspjehom i invazivnošću, proučavao sam ponašanja vrste
Podarcis siculus u tri ekološka konteksta: 1) na interspecijskoj razini sa subdominantnom
vrstom Podarcis melisellensis; 2) na intraspecijskoj razini između populacija u različitim
okolišima; i 3) između populacija u sličnim okolišima. Podarcis siculus pokazala je veću
razinu istraživačkog ponašanja i nižu razinu odvažnosti te je konzumirala više hrane od P.
melisellensis. Populacija P. siculus na otočiću s ograničenim resursima pokazala je nižu
razinu agresivnosti i pratećih svojstva nego populacija s kopna. Populacije iz sličnih okoliša
su pokazale nisku diferencijaciju u ponašanju. Međutim, zabilježene su značajne razlike u
bihevioralnim osobinama između jedinki, a određene bihevioralne osobine su bile međusobno
korelirane. Općenito, specifična ponašanja P. siculus te intra- i inter-populacijska varijabilnost
u ponašanju mogu pomoći u kompeticiji i invazivnosti
Quantitative determination of neonicotinoides and its metabolites in honeybees and honeybee products using liquid chromatography-tandem mass spectrometry
No full textRazvijena je metoda za kvantitativno određivanje neonikotinoida i njihovih metabolita u
uzorcima pčela i pčelinjim proizvodima vezanim sustavom tekućinska kromatografija –
tandemna spektrometrija masa. Predloženi analitički postupci validirani su u skladu sa
smjernicama SANTE/12682/2019 ispitivanjem radnih značajki metode u svim ciljanim
matricama. Preciznost, izražena kao relativna standardna devijacija, bila je < 20 %, dok je
točnost, izražena kao analitički povrat bila u rasponu od 70 % do 120 %, za većinu
neonikotinoida i njihovih metabolita. U okviru validacije metoda procijenjena mjerna
nesigurnost iznosila je < 50 % za sve analite. Predloženom metodom analizirano je ukupno
664 uzoraka pčela i pčelinjih proizvoda uzorkovanih tijekom 2019. i 2020. godine u svim
regijama Republike Hrvatske. Ostatci neonikotinoida i njihovih metabolita detektirani su u
17,6 % uzoraka, a tiakloprid je bio najčešće prisutan analit u obje godine istraživanja.A method for the quantitative determination of neonicotinoids and their metabolites in
honeybee and bee product samples using the liquid chromatography - tandem mass
spectrometry system was developed. The proposed analytical procedures were validated in
accordance with the SANTE/12682/2019 guidelines by evaluating the performance
characteristics of the method in all target matrices. Precision, expressed as relative standard
deviation, was < 20 %, while accuracy, expressed as analytical recovery, ranged from 70 %
to 120 %, for most neonicotinoids and their metabolites. As part of method validation, the
estimated measurement uncertainty was < 50 % for all analytes. A total of 664 samples of
honeybees and bee products, sampled during 2019 and 2020 in all regions of the Republic of
Croatia, were analysed using the proposed method. Residues of neonicotinoids and their
metabolites were detected in 17.6 % of samples, and thiacloprid was the most frequently
present analyte in both years of the study
Osnovno stanje vala gustoće naboja u interkaliranom grafitu CaC6
No full textThe physics of graphite intercalated compounds (GIC), in spite of the fact that these materials were synthesised quite long ago, is still the focus of modern theoretical and experimental research. The interest of the scientific community is focused on GIC, mostly due to the possibility of forming unusual superconducting (SC) and charge density wave (CDW) ground states. We focus on one of these materials, CaC6, in which experiments clearly demonstrate the existence of CDWs. The very existence of the CDW is quite unusual due to the fact that the graphene Fermi surface, chemically doped by calcium intercalates, shows no so-called "nesting property," i.e., no Peierls mechanism, which has been the paradigmatic mechanism behind the CDW instability since the 1950s and was experimentally realised during the 1980s in highly anisotropic materials with open Fermi surfaces. We show that the CaC6 Fermi surface undergoes a different CDW stabilisation mechanism based on the topological reconstruction (of the Fermi Surface) and critical strength of the electron-phonon interaction. Such topological reconstruction, which transforms the set of closed Fermi pockets into open Fermi sheets, has profound effects on transport and magnetotransport properties. In this thesis, we show how the Fermi surface reconstruction stabilises the CDW ground state. Furthermore, we present the magnetotransport properties, taking into account magnetic breakdown.Cilj ovog rada je opisati i objasniti pojavu nastanka valova gustoće naboja u interkaliranom grafitu CaC6 tzv. minimalnim modelom, koji sadrži samo fizikalne odlike nužne za samu egzistenciju pojave vala gustoće naboja. Inicijalni problem sustava vezanih elektrona i fonona opisujemo dvodimenzionalnim Fröhlichovim hamiltonijanom. Njega tretiramo u aproksimaciji srednjeg polja i uvodimo kompleksni parametar uređenja. Nakon tretmana aproksimacijom srednjeg polja dobivamo izraz za elektronske vrpce. Kako bismo dobili oblik rekonstruiranih elektronskih vrpci, posežemo za posebno razvijenim metodama razvoja podintegralnih funkcija kako bismo dobili analitički rezultat za kondenzacijsku energiju sustava. Da bi novonastalo stanje s valom gustoće naboja bilo stabilno, kondenzacijska energija mora biti pozitivna, a njezina maksimizacija vrši se optimizacijom izraza po samosuglasnom valnom vektoru vala gustoće i po njegovom parametru uređenja. Nadalje, određujemo koji je fononski mod kristalne rešetke u CaC6 strukturi zaslužan za pojavu vala gustoće. Pokazuje se da je to TA mod titranja kalcijeve superrešetke u njezinoj ravnini. Ovaj mod je "najmekši", odnosno ima najmanju energiju, ali i najjače vezan je vezan na grafenske elektrone čime su eliminirani svi grafenski vibracijski modovi, kao i kalcijevi fleksuralni modovi titranja