UiS Brage
Not a member yet
19747 research outputs found
Sort by
Uncertainty Assessment and Inference for VLAD plots
Full text linkThe main subject of this thesis is to explore the uncertainty assessment of VLAD plots ("Variable-Life-Adjusted-Display" plots). VLAD plots are statistical plots that plot the cumulative sum of the expected value minus the observed value. The main focus of this problem is to explore how to create control limits and confidence intervals for VLAD-plots since there are no established methods yet. Some proposed methods explored earlier by others to create control limits for VLAD are using CUSUM based on log-likelihood ratio, and EWMA (Exponentially Weighted Mean Average). These methods, along with a new method created specifically for this thesis called MP (Moving Proportion), will be used in this thesis to create control limits. All of these methods were chosen since they are capable of detecting small shifts within a dataset. The data within the HBB dataset became the basis of how these methods work in a real-life setting. The HBB dataset contains the birth outcomes from a hospital in Tanzania, along with other relevant statistical information. The dataset was used for a project with the same name. In addition to looking at the properties of VLAD plots in a real-life setting, the properties of VLAD plots on simulated data have also been considered. One of the important results in this thesis is that all three methods are capable of creating control limits for VLAD-plots where the signal points in the VLAD-plots match the signal points in the respective plot for the used processing method, for example if one used CUSUM to create control limits for VLAD then the signal points on the VLAD-plot would match the signal points on the CUSUM-plot. The method that seems to work most consistently in the settings in this thesis was CUSUM based on the log-likelihood ratio. Another important result is that there exist three types of confidence intervals for VLAD-plots depending on what variance one is interested in, type 1 only considers the variance in the observed values, type 2 considers only the variance in the expected values while keeping the observed values fixed and type 3 considers the variance in both expected and observed values.The main subject of this thesis is to explore the uncertainty assessment of VLAD plots ("Variable-Life-Adjusted-Display" plots). VLAD plots are statistical plots that plot the cumulative sum of the expected value minus the observed value. The main focus of this problem is to explore how to create control limits and confidence intervals for VLAD-plots since there are no established methods yet. Some proposed methods explored earlier by others to create control limits for VLAD are using CUSUM based on log-likelihood ratio, and EWMA (Exponentially Weighted Mean Average). These methods, along with a new method created specifically for this thesis called MP (Moving Proportion), will be used in this thesis to create control limits. All of these methods were chosen since they are capable of detecting small shifts within a dataset. The data within the HBB dataset became the basis of how these methods work in a real-life setting. The HBB dataset contains the birth outcomes from a hospital in Tanzania, along with other relevant statistical information. The dataset was used for a project with the same name. In addition to looking at the properties of VLAD plots in a real-life setting, the properties of VLAD plots on simulated data have also been considered. One of the important results in this thesis is that all three methods are capable of creating control limits for VLAD-plots where the signal points in the VLAD-plots match the signal points in the respective plot for the used processing method, for example if one used CUSUM to create control limits for VLAD then the signal points on the VLAD-plot would match the signal points on the CUSUM-plot. The method that seems to work most consistently in the settings in this thesis was CUSUM based on the log-likelihood ratio. Another important result is that there exist three types of confidence intervals for VLAD-plots depending on what variance one is interested in, type 1 only considers the variance in the observed values, type 2 considers only the variance in the expected values while keeping the observed values fixed and type 3 considers the variance in both expected and observed values
Fact Checking TikTok
Full text linkThe rapid growth of short-form video platforms such as TikTok introduces complex chal-
lenges for misinformation detection due to their highly multimodal, dynamic, and often noisy
content. This thesis takes an exploratory approach to address these challenges by develop-
ing and evaluating a modular pipeline designed to identify potentially checkworthy content.
The goal is to investigate which techniques contribute meaningfully to detecting misleading
or significant content in short-form video.
The pipeline integrates components across multiple modalities—including OCR, object
detection, deepfake detection, audio transcription, visual captioning, and semantic summa-
rization and classification. Outputs from each module are combined through a rule-based
verdict system to assign one of three labels: Checkworthy, Not_Checkworthy, or Advertise-
ment.
Evaluation was conducted on a manually annotated dataset of 249 TikTok videos. Find-
ings indicate that the transcript-based module had the strongest influence on verdict assign-
ment, while others, such as deepfake and weapon detection, were constrained by limited data
representation. Importantly, all results were derived from out-of-distribution data to assess
generalizability rather than benchmark performance.
This work does not seek to present a complete or optimized pipeline but rather to explore
and assess modular strategies for future refinement. The modular structure proved beneficial
for isolating errors and adapting to the noisy, diverse nature of real-world content. These
findings lay the groundwork for further research into scalable, domain-agnostic approaches
to video-based misinformation detection.The rapid growth of short-form video platforms such as TikTok introduces complex chal-
lenges for misinformation detection due to their highly multimodal, dynamic, and often noisy
content. This thesis takes an exploratory approach to address these challenges by develop-
ing and evaluating a modular pipeline designed to identify potentially checkworthy content.
The goal is to investigate which techniques contribute meaningfully to detecting misleading
or significant content in short-form video.
The pipeline integrates components across multiple modalities—including OCR, object
detection, deepfake detection, audio transcription, visual captioning, and semantic summa-
rization and classification. Outputs from each module are combined through a rule-based
verdict system to assign one of three labels: Checkworthy, Not_Checkworthy, or Advertise-
ment.
Evaluation was conducted on a manually annotated dataset of 249 TikTok videos. Find-
ings indicate that the transcript-based module had the strongest influence on verdict assign-
ment, while others, such as deepfake and weapon detection, were constrained by limited data
representation. Importantly, all results were derived from out-of-distribution data to assess
generalizability rather than benchmark performance.
This work does not seek to present a complete or optimized pipeline but rather to explore
and assess modular strategies for future refinement. The modular structure proved beneficial
for isolating errors and adapting to the noisy, diverse nature of real-world content. These
findings lay the groundwork for further research into scalable, domain-agnostic approaches
to video-based misinformation detection
Impact of On-Demand Electricity Services on Household Consumer Behavior in European Countries
No full text
Gartnernæringen i endring: En SWOT-analyse av hvordan nye miljøkrav påvirker norsk veksthusproduksjon.
No full textDenne bacheloroppgaven undersøker hvordan skjerpede miljøkrav påvirker gartnernæringen i Norge. Vi har valgt å benytte en SWOT-analyse for å belyse hvilke styrker, svakheter, muligheter og trusler som miljøkravene medfører, og har kombinert dette med en kvalitativ tilnærming gjennom intervjuer med tre ulike gartnere fra samme region. Bakgrunnen for temaet er det økende fokuset på klima og bærekraft, både fra myndigheter og samfunnet generelt.
Gjennom teori, blant annet Verdenskommisjonen for miljø og utvikling (1987) sin definisjon av bærekraft og Rogers (2003) teori om innovasjon, og praktiske eksempler fra virkelige aktører i bransjen, har vi fått innsikt i hvordan miljøkravene oppleves og håndteres i praksis. Det kommer frem at mange produsenter ser miljøkrav som nødvendige, men også utfordrende. Spesielt små aktører opplever presset som stort, blant annet grunnet høye investeringskostnader og uforutsigbare støtteordninger.
Samtidig er det flere gartnere som peker på mulighetene som ligger i en grønnere profil. Det gjelder alt fra forbedret omdømme, nye markedsmuligheter og effektivisering av drift – til mulig støtte fra myndighetene. Men for at dette skal bli en realitet, må det skapes bedre rammebetingelser og forutsigbarhet.
Oppgaven konkluderer med at miljøkravene ikke nødvendigvis er en trussel for gartnernæringen, men at de krever tilrettelegging og støtte for å kunne fungere som en positiv drivkraft. Det grønne skiftet kan bli en mulighet for vekst og utvikling, dersom næringen og myndighetene samarbeider godt. Gartnernæringen står overfor viktige endringer, og hvordan disse håndteres vil kunne påvirke bransjens utvikling fremover
Designforslag for Randaberg torg
No full textDenne bacheloroppgaven tar for seg hvordan Randaberg torg kan forbedres som offentlig byrom for å styrke områdets identitet og legge bedre til rette for opphold, aktivitet og sosialt liv. Torget fungerer i dag som en gjennomfartsåre med parkeringsarealer og begrensede fasiliteter for ulike brukergrupper, noe som bidrar til lite opphold og bruk. Samtidig har torget et stort potensial som møteplass i sentrum av Randaberg.
Gjennom en grundig stedsanalyse, inspirert av Stavanger kommunes veileder, identifiseres både styrker og svakheter ved dagens utforming. Arbeidet bygger videre på teorier om byrom, stedsidentitet og inkluderende design, med særlig vekt på Jan Gehls 12 kvalitetskriterier for gode byrom. Videre er det innhentet inspirasjon fra referanseprosjekter i Sandnes og Ålgård.
Målet med oppgaven er å foreslå konkrete tiltak som kan gjøre torget mer funksjonelt, attraktivt og inkluderende. Dette innebærer blant annet å skape tydeligere ganglinjer, bedre sittemuligheter, oppgraderte lekearealer, mer vegetasjon, samt universell utforming som ivaretar behovene til ulike aldersgrupper og personer med funksjonsnedsettelser.
Oppgaven konkluderer med et helhetlig designforslag som tar utgangspunkt i Randabergs kulturarv og visjonen om "den grønne landsbyen". Forslaget har som mål å gjøre torget til et levende og inviterende byrom som kan bidra til økt trivsel, lokal handel og sosialt fellesskap i Randaberg kommune.This bachelor's thesis addresses how the square in Randaberg can be improved as a public urban space to strengthen the area's identity and better facilitate for dwelling, activity and social life. The square currently functions as a thoroughfare with parking areas and limited facilities for various user groups, which contributes to low occupancy and use. At the same time, the square has great potential as a meeting place in the center of Randaberg.
Through a thorough site analysis, inspired by the Stavanger Municipality's supervisor, both strengths and weaknesses of the current design are identified. The work builds on theories about urban space, place identity and inclusive design, with particular emphasis on Jan Gehl's 12 quality criteria for good urban spaces. Furthermore, inspiration has been obtained from reference projects in Sandnes and Ålgård.
The goal of the thesis is to propose concrete measures that can make the square more functional, attractive and inclusive. This includes creating clearer walkways, better seating options, upgraded play areas, more vegetation, and universal design that meets the needs of different age groups and people with disabilities.
The task concludes with a comprehensive design proposal that is based on Randaberg's cultural heritage and the vision of the "green village". The proposal aims to make the square a vibrant and inviting urban space that can contribute to increased well-being, local commerce, and social community in Randaberg municipality
Comparative Study of Transformer & LSTM Architectures for Anomaly Detection in Multivariate Time Series Data
Full text linkAnomaly detection of multivariate time series data is critical to various industries
as it may contain information about the fault in the system or a security breach.
Using Long Short-Term Memory (LSTM) networks has long been the default approach
for tackling this modeling challenge. Recent Transformer architectures,
however, may deliver better performance. The transformer model and the standard
LSTM model using RNN architecture are compared through this study to
detect anomalies in multivariate time series data. The Jena Climate dataset was
used to train both models on the temperature, humidity and air density values to
make future predictions and find anomaly from the prediction error. Both models
are trained to be next-step forecasters. When the reconstruction error is above
an error high threshold, a flag is raised. The LSTM has slightly lower point-wise
error than the Transformer, but not necessarily better trends. Most importantly,
both designs to identify almost all injected anomalies. The LSTM is more practical
for edge deployment because it has far fewer parameters and trains faster. But
when global structures play a more important role than resources, Transformers
are worth a look.Anomaly detection of multivariate time series data is critical to various industries
as it may contain information about the fault in the system or a security breach.
Using Long Short-Term Memory (LSTM) networks has long been the default approach
for tackling this modeling challenge. Recent Transformer architectures,
however, may deliver better performance. The transformer model and the standard
LSTM model using RNN architecture are compared through this study to
detect anomalies in multivariate time series data. The Jena Climate dataset was
used to train both models on the temperature, humidity and air density values to
make future predictions and find anomaly from the prediction error. Both models
are trained to be next-step forecasters. When the reconstruction error is above
an error high threshold, a flag is raised. The LSTM has slightly lower point-wise
error than the Transformer, but not necessarily better trends. Most importantly,
both designs to identify almost all injected anomalies. The LSTM is more practical
for edge deployment because it has far fewer parameters and trains faster. But
when global structures play a more important role than resources, Transformers
are worth a look
Forming som arena for bærekraftig utvikling
Full text linkI denne bachelor oppgaven har jeg sett på problemstillingen hvordan formings aktiverer kan
fremme bærekraftig utvikling. for å belyse dette teamet gjennomføre jeg intervjue med
barnehagelærer og en didaktisk formings aktivitet med fokus på bærekraftig utvikling etter
gjennomført formingsaktivitet gjennomførte jeg et oppfølgings intervju med barnehagelærer
for å høre hennes tanker og refleksjoner av en formings aktivitet med et bærekraftig fokus.
Funnene mine viser at bærekraftig utvikling kan fremmes gjennom formingsaktiviteter og
bærekraftige materialer kan berike formingspraksisen i barnehagen. Funnene mine viser også
viktigheten av barnehagelæreren inn i arbeidet og jeg ser på hvordan barnehagelærers
kunnskaper og interesse kan påvirke arbeidet med bærekraftig utvikling. jeg har også sett på
hva natur og gjenbruks materialer er og hvordan de kan påvirke formingsaktiviteter i
barnehagen.In this bachelor's thesis, I have looked at the problem of how forming activates can
promote sustainable development. to shed light on this team I conduct interviews with
kindergarten teacher and a didactic training activity with a focus on sustainable development after
completed training activity, I conducted a follow-up interview with the kindergarten teacher
to hear her thoughts and reflections on a shaping activity with a sustainable focus.
My findings show that sustainable development can be promoted through shaping activities and
sustainable materials can enrich the shaping practice in the nursery school. My findings also show
the importance of the kindergarten teacher into the work and I look at how kindergarten teachers
knowledge and interest can influence work on sustainable development. I have also watched
what nature and recycled materials are and how they can affect shaping activities i
the kindergarten
Kvinners seksuelle helse etter brystkreftbehandling
Full text linkSammendrag
Bakgrunn
Brystkreft er den hyppigst forekommende kreftformen blant kvinner og rammer årlig rundt 4000 kvinner i Norge årlig (Kreftforeningen, & Brystkreftforeningen, n.d). Diagnosen innebærer et krevende behandlingsforløp, hvor utløst eller fremskyndet klimakterium ofte følger behandlingen og kan påvirke kvinnens seksuelle helse. Fysiske plager som tørre slimhinner, redusert elastisitet, nedsatt libido og smerter ved samleie er vanlige symptomer, og disse opptrer ofte i kombinasjon med følelsesmessige endringer forårsaket av sykdom og behandling. Til tross for at slike plager er utbredte, viser både nasjonal og internasjonal forskning at helsepersonell i liten grad kommuniserer tilstrekkelig om seksuell helse med pasienter.
Hensikt
Hensikten med oppgaven er å belyse hvordan sykepleier kan støtte kvinner i å ivareta deres seksuelle helse etter brystkreftbehandling.
Metode
Vi har i denne bacheloroppgaven valgt integrativ litteraturoversikt som metode. Seks forskningsartikler ble utvalgt og analysert, og oppgaven bygger på sentrale funn fra disse artiklene, i tillegg til relevant faglitteratur, sykepleierteori og egne refleksjoner.
Resultat
Det er betydelige utfordringer rundt kommunikasjon om seksuell helse og støtte for intime utfordringer hos brystkreftpasienter. Felles utfordringer inkluderer manglende kommunikasjon og støtte fra sykepleiere, tidspress og manglende rutiner, kulturelle barrierer og behov for bedre opplæring og organisatorisk støtte. Integrering av seksuell helse som en del av rutinemessig omsorg kan forbedre pasientopplevelsen og kvaliteten på omsorgen som tilbys.
Nøkkelord: Sykepleie, sykepleie funksjon, seksuell helse, brystkreft, holdninger, kompetanse, kommunikasjon, informasjon, tid, rutiner, barrierer, kulturelle normer, stigma
