240106 research outputs found
Sort by
Clickstreams personified: A study on Web Usage Mining for generating personas.
No full textAs the world changes, so must the persona litteratureby considering the rising interrest in big data. This project aims to contribute to this, by an exploratory studyof the use of Web Uage Mining based personas in anexisting product design setting. To do this, the methodis compared to a traditional persona generating processby executing both processes on the same user group. Results indicate a need to improve the current Web UsageMining method to achieve af fully automatic personageneration. As for the traditional persona generatingprocedure, a need for a complete guide that takes contextualized persona template design into consideration isargured. To explore both methods relevance, resultingpersonas are used under same design task by productteam members. Results support the continued relevanceof traditional personas, and show Web Usage Miningbased personas as relevant supplementary data, but notas a suitable alternative, as lack of personal informationand effective data presentation is problematic. Futurestudies should aim at improving Web Usage Miningpersona automation, and further explore relevant methods for including user frustrations and effective datavisualisation techniques.<br/
Hyperbolic Generalized Category Discovery: Hyperbolic visual learning and clustering for generalized category discovery
No full textThis thesis investigates the use of hyperbolic visual learning in Generalized Category Discovery (GCD), an open-world classification problem in which a model is tasked with classifying both known and novel classes.To that end, two GCD methods are adapted to learn embeddings in the Lorentz model of hyperbolic geometry. Furthermore, the K-Means algorithm is modified to perform non-parametric clustering on hyperbolic embeddings.The experiments showcase that the non-parametric method benefits from learning in Lorentz space, while the parametric method does not, with its best results in Euclidean geometry.Furthermore, experiments with the hyperbolic K-Means demonstrate increased accuracy when clustering embeddings directly in Lorentz space. However, K-Means in the Poincaré model of hyperbolic geometry suffers from instability, failing to generate valid clusters depending on the initialization of the cluster prototypes
EnergyBench: A Holistic and Systematic Benchmark for Measuring the Correctness and Energy-Efficiency of LLM-Generated Code
No full textThe growing adoption of Artificial Intelligence (AI) is a primary contributor to increased energy demands in data centers, raising environmental concerns for the Information, Communication, and Technology (ICT) sector as a whole. In the last five years, advances in areas like natural language processing and computer vision have significantly increased the global adoption of AI. An important topic of research lies at the intersection of Large Language Models (LLMs) with modern software development. The software industry is currently facing a steady transformation towards automated code generation, with LLM assistants and LLM-powered code editors at the forefront. The negative environmental impact of running these large AI models is certain, but questions still remain unanswered about the sustainability of LLM-generated code. Current benchmarks show that LLMs have the ability to generate energy-efficient code when given optimization prompts. However, the results remain unclear when arguing for which problems, programming languages, and prompting strategies lead to the best trade-off between accuracy and energy efficiency. To address these gaps, EnergyBench is introduced—a benchmarking framework that takes a holistic and systematic approach to analyzing the elements that most impact the ability of LLMs to generate correct and energy-efficient code. Experiments show that prompts focused on reducing two key performance metrics make code more energy-efficient by as much as 91.9% in 5 out of the 7 tested LLMs, though this comes at the cost of lowered overall accuracy. Additional experiments show that the systematic approach EnergyBench takes reveals gaps unexplored by the current state-of-the-art. LLMs are highly sensitive to prompt contents: two tweaks made when defining a programming problem in an LLM prompt have drastic and opposite effects on both accuracy and energy efficiency. In one case, efficiency is increased by more than 4×, while in another, accuracy drops to zero. EnergyBench's implementation is released to the public, inviting the community to contribute to the existing set of tests in the hopes that a broader, more detailed view of the sustainability of LLM coding is reached
A field that demands change. An empirical study of Danish public school teachers' experience in a politically changeable work environment
No full textAnimated Realities: The Synaesthesia Experience (A.R.T.E.):Part three – What is it like to experience Synesthesia?
No full textIn this third of three short movies in the Animated Realities: The Synaesthesia Experience (A.R.T.E.) project three different synaesthetes describe how they experience their specific synaesthesia
The crisis that normalised time-shifting:Energy flexibility, price awareness and care during the energy crisis in Denmark
Get PDFThe energy crisis during Winter 2022–2023 placed energy high on the public agenda in Denmark, including a hitherto unseen focus on dynamic electricity prices. This paper identifies a normalisation of following electricity prices online during the crisis and changing practices accordingly, the aim of the paper being to learn from this normalisation theoretically as well as policy-wise. The paper is based on a quantitative survey (N = 1000) and in-depth qualitative interviews (n = 30). Results showed that among those who indicate that they have flexible electricity prices almost 70% say that they follow prices daily, and an even larger share indicate that they are more aware of the timing of their electricity use compared to one year ago. Furthermore, more than half of the households used timers on their washing machines and dishwashers to time-shift appliance-use. Through an analysis of variation in interest and practices of time-shifting, we construct a parameter that we call ‘care for the energy system’, which showed more correlations than households’ socio-economy. The qualitative interviews revealed how experiences and engagement in time-shifting of practices for some households was prompted by tight economic constraints and for other households by care for energy and interest in green energy consumption. The results suggest that variable prices coupled with media and authority communication can encourage people to care for the energy system through time-shifting. However, as economic incentives can have social consequences, price signals should be coupled with financial support for vulnerable households.</p
A Game-Theoretic Perspective for Efficient Modern Random Access
No full textModern random access mechanisms combine packet repetitions with multi-user detection mechanisms at the receiver to maximize the throughput and reliability in massive Internet of Things (IoT) scenarios. However, optimizing the access policy, which selects the number of repetitions, is a complicated problem, and failing to do so can lead to an inefficient use of resources and, potentially, to an increased congestion. In this paper, we follow a game-theoretic approach for optimizing the access policies of selfish users in modern random access mechanisms. Our goal is to find adequate values for the rewards given after a success to achieve a Nash equilibrium (NE) that optimizes the throughput of the system while considering the cost of transmission. Our results show that a mixed strategy, where repetitions are selected according to the irregular repetition slotted ALOHA (IRSA) protocol, attains a NE that maximizes the throughput in the special case with two users. In this scenario, our method increases the throughput by 30% when compared to framed ALOHA. Furthermore, we present three methods to attain a NE with near-optimal throughput for general modern random access scenarios, which exceed the throughput of framed ALOHA by up to 34%
Spatial Characterization of Indoor Radio Channel in the D-Band at 165 GHz
No full textThis paper presents an experimental study of the indoor radio channel characteristics at 165 GHz in the D-band. Using wideband measurements based on sliding correlation technique (swept time delay cross-correlation, STDCC) with a pseudo-random binary sequence (PRBS) of length 223, we analyze key parameters such as the power delay profile, delay spread, and fading effects as the receiver moves along a linear path. The results indicate significant multipath propagation, with increasing delay spread corresponding to distance. Additionally, signal quality degradation metrics, including Error Vector Magnitude (EVM) and Modulation Error Ratio (MER) are evaluated. The obtained results are consistent with an indoor scenario dominated by a LOS component, with no significant presence of multipath contributions
P1 Anmelderne: Program om dokumentarfilmen Vini Jr. på Netflix
No full textDen brasilianske fodboldspiller Vinicius Jr. er personificeringen af drengedrømmens indfrielse og sportsverdenens magi. En fattig knægt fra Rio de Janeiros favela, der spiller gadefodbold i bare tæer, bliver opdaget, oplært og solgt til selveste Real Madrid.Med sine legesyge driblinger sidder millioner af fans forventningsfuldt cleanet til skærmen, hver gang Vinicius Jr. tørner ud for den spanske kongeklub. Men brasilianeren er også en kontroversiel skikkelse: Provokerende over for modspillere og dommere, og et offer for den racisme, som fortsat hærger spansk fodbold.Det hele forsøges skildret i Netflix' nye dokumentarfilm, "Vini Jr.". Men er dokumentaren mere en fan-film end et objektivt portræt? Hvad fortæller den om den måde, som unge mænd skal håndtere svimlende opmærksomhed og lige så store pengesummer? Og hvorfor er det så svært at komme racisme i fodbold til livs