BIBSYS: Open Journals Systems
Not a member yet
393 research outputs found
Sort by
Stochastic Local Search Heuristics for Efficient Feature Selection: An Experimental Study
Get PDFFeature engineering, including feature selection, plays a key role in data science, knowledge discovery, machine learning, and statistics. Recently, much progress has been made in increasing the accuracy of machine learning for complex problems. In part, this is due to improvements in feature engineering, for example by means of deep learning or feature selection. This progress has, to a large extent, come at the cost of dramatic and perhaps unsustainable increases in the computational resources used. Consequently, there is now a need to emphasize not only accuracy but also computational cost in research on and applications of machine learning including feature selection. With a focus on both the accuracy and computational cost of feature selection, we study stochastic local search (SLS) methods when applied to feature selection in this paper. With an eye to containing computational cost, we consider an SLS method for efficient feature selection, SLS4FS. SLS4FS is an amalgamation of several heuristics, including filter and wrapper methods, controlled by hyperparameters. While SLS4FS admits, for certain hyperparameter settings, analysis by means of homogeneous Markov chains, our focus is on experiments with several realworld datasets in this paper. Our experimental study suggests that SLS4FS is competitive with several existing methods, and is useful in settings where one wants to control the computational cost
Digital carbon footprint awareness among digital natives: an exploratory study
Get PDFChanges in use practices due to COVID-19 have illustrated the potential of digital technology as a catalyst of more sustainable and pro-environmental behavior. At the same time, the energy consumption and environmental impact of digital applications and services has been put more firmly on the agenda. In this paper, we adopt a bottom-up approach to explore digital natives’ awareness of their digital carbon footprint, i.e., related to their use of digital services and applications. We present findings from 21 semi-structured in-depth interviews that were conducted to explore (1) to which extent digital natives are aware of and reflect on their digital carbon footprint, (2) what could motivate efforts to reduce this footprint and (3) which compromises they might be willing to make in this respect. The findings point to low awareness of the carbon footprint of digital applications and services. The lack of technological understanding, public information and social awareness about the topic were identified as important factors. In terms of the motivation for adopting pro-environmental digital habits, we found that several factors indirectly contribute to this goal, including the striving for personal wellbeing. Finally, the results indicate some willingness to change and make compromises, albeit not an unconditional one: the alignment with other goals (e.g., personal well-being) and nature of the perceived sacrifice and its impact play a key role. With this work, we aim to strengthen ongoing efforts to increase users’ awareness and to stimulate more sustainable and well-being supporting digital consumption
Who uses VR in Norway?
Get PDFDespite much hype, the extent to which VR is accepted and used by the general population is unclear. To shed light on this, we conducted a representative panel survey with 936 respondents in the spring of 2021. The survey revealed that while many know what VR is, only 20 % have tried it, and only 0.6 % use VR more than once a week. This poster explores the demographics of VR users in Norway
QuickFeed on Programming Assignments
Get PDFIn the last decade, programming has become an increasingly important tool for almost all science and engineering disciplines. To this end, programming exercises have become an essential tool for students to learn the craft of programming and apply, model, and evaluate other scientific techniques
Brain Tumour Classification using Convolutional Neural Network
Get PDFIn this extended abstract, we address the problem of classifying MRI images of di?erent brain tumours to facilitate the development of an automated system for early brain cancer detection. In particular, we adapt the convolutional neural network based on the AlexNet architecture to develop a model that achieves a classi?cation accuracy of 97.5% on average on a real-world dataset containing MRI images of healthy brains and three di?erent kinds of tumours
Digital Platform Ecosystem Governance: Preliminary Findings and Research Agenda
Get PDFThis paper explores collaborative governance in digital platform ecosystems and the governance challenges that may occur in such environments. We analyze three different digital platform ecosystems and identify six unresolved key governance issues that we believe are central to the type of digital platform ecosystems we address. This paper has three contributions. First, we add to the literature on digital platform ecosystems by revealing a set of governance challenges regarding ecosystem forming and sustainability. Second, our findings may serve as recommendations for organizations that are planning to establish or that are already running an ecosystem based on a digital platform. Third, we contribute to digital platform ecosystem research by proposing an agenda for future research in this area
Hva skjer når etablerte, samlokaliserte agile team må arbeide og samarbeide fra hjemmekontor?
Get PDFI denne artikkelen utforsker vi hvordan etablerte, samlokaliserte agile team påvirkes av overgangen til en virtuell arbeidsform. For å svare på spørsmålene har vi studert tre agile team i et norsk utviklingsselskap, som på grunn av Covid-19 ble nødt til å bytte arbeidssted fra et felles kontorlandskap til individuelle hjemmekontor. Studien finner at virtuelt arbeid påvirker den agile arbeidsformen ved at det blir færre interaksjoner, mer skriftlig kommunikasjon, mer formaliserte relasjoner, samt økt bruk av dokumentasjon. Videre finner vi at overgangen til hjemmekontor fikk konsekvenserfor teamledelsen og en annen lederstil enn hva tradisjonelle, samlokaliserte agile team har behov for. Studien gir tre bidrag. For det første bidrar den til litteraturen om agil metode ved å dokumentere hva som skjer når teammedlemmene skal arbeide og samarbeide fra hjemmekontor. For det andre bidrar studien til litteraturen om hjemmekontor, og effektene av hjemmekontor på samarbeid. Og for det tredje bidrar studien til praksis ved at det foreslås tiltak som kan bøte på ulempene ved at teammedlemmene ikke kan oppholde seg på samme arbeidssted
Peer Code Review as Formative Assessment: A Case Study from a Database Course Project
Get PDFCode review is an important quality assurance activity for software engineers. In addition, both reviewers and developers may improve their professional competencies when participating in code reviews. This paper describes a three-week database course project where the students designed and developed a database and a database application. After a three-week project work period, the students submitted their designs and their code repositories for peer review. 115 students participated in the code review of the 68 repositories submitted for review. The students could make use of the feedback they received and what they learned from studying other student’s code when revising the project for the final submission, which was graded by the teaching staff.
In their final reports, the students reported on benefits and barriers for effective peer review that seemed to be more accentuated when the purpose was formative assessment. Benefits included high level thinking and deeper levels of learning and giving the students different points of view when reviewing their own solutions. Some of the barriers were the students’ lack of domain knowledge and lack of confidence in being reviewers, the quality of the projects submitted for review, the review workload, and the amount of work needed to modify the code after the review.
 
“Best Practice” without Evidence – Agile Software Methodology as Example
Get PDFDespite the essentiality of education, and the widely known unscientific nature of expert opinion, education in general appears to be based on expert opinion. The example analyzed herein is of Agile software methodology, which is deemed a best practice and therefore taught in most IT studies, in Norway and most probably internationally. This is despite that it appears to be a well known fact within its respective field that the Agile methodology lacks scientific justification. A tertiary analysis was conducted to test this well known fact and to serve as basis for exploring what should be considered sufficient evidence for inclusion within official education. The result of the tertiary study is that, indeed, the evidence for the Agile methodology is scarce at best. A method to avoid such mistake is suggested, which could be valuable to science in general. This method entails employing philosophers of science, epistemologists, to counteract potential expert biases and verify the curriculum before it is accepted in official education