Tind Technologies (Norway)
Hes-so: ArODES Open Archive (University of Applied Sciences and Arts Western Switzerland / Haute école spécialisée de Suisse occidentale / FH Westschweiz)Not a member yet
15764 research outputs found
Sort by
Discrete graph auto-encoder
No full textDespite advances in generative methods, accurately modeling the distribution of graphs remains a challenging task primarily because of the absence of predefined or inherent unique graph representation. Two main strategies have emerged to tackle this issue: 1) restricting the number of possible representations by sorting the nodes, or 2) using permutation-invariant/equivariant functions, specifically Graph Neural Networks (GNNs). In this paper, we introduce a new framework named Discrete Graph Auto-Encoder (DGAE), which leverages the strengths of both strategies and mitigate their respective limitations. In essence, we propose a strategy in 2 steps. We first use a permutation-equivariant auto-encoder to convert graphs into sets of discrete latent node representations, each node being represented by a sequence of quantized vectors. In the second step, we sort the sets of discrete latent representations and learn their distribution with a specifically designed auto-regressive model based on the Transformer architecture. Through multiple experimental evaluations, we demonstrate the competitive performances of our model in comparison to the existing state-of-the-art across various datasets. Various ablation studies support the interest of our method
Can linear algebra create perfect knockoffs ?
No full textAs new Model-X knockoff construction techniques are developed, primarily concerned with determining the correct conditional distribution from which to sample, we focus less on deriving the correct multivariate distribution and instead ask if “perfect” knockoffs can be constructed using linear algebra. Using mean absolute correlation between knockoffs and features as a measure of quality, we produce knockoffs that are pseudo-perfect, however, the optimization algorithm is computationally very expensive. We outline a series of methods to significantly reduce the computation time of the algorithm
Introduction to the special issue on realistic synthetic data ::generation, learning, evaluation
No full text
High-speed AI image space wavefront sensing using embedded computing ::achieving 1000 frames per second
No full textAdaptive optics (AO) is a crucial field in optics, requiring precise wavefront sensing. This paper presents a fast wavefront sensing solution based on AI4Wave, a patented technology for AI-based image-space wavefront sensing. AI4Wave eliminates dedicated sensors like Shack-Hartmann, using only a camera. This high-speed implementation leverages AI edge computing on an NVIDIA Jetson module. The system captures defocused images processed by a feedforward neural network (NN) trained exclusively on synthetic data. This enables real-time phase retrieval, overcoming Shack-Hartmann’s limitations and handling large wavefront errors. AI4Wave employs synthetic, normalized data, making it robust to optical layout changes. Using NVIDIA TensorRT and advanced AI edge computing, it achieves processing speeds of 1000 frames per second, supporting various optical setups. The deterministic feedforward NN approach ensures consistent results without iterative optimization. Preliminary tests show high accuracy and repeatability, with exposure times as short as 24 µs, capturing environmental perturbations. This provides a reliable, industrial-grade solution for high-speed wavefront sensing
Nurses’ clinical practice in nursing homes ::depressive symptoms and fall risk assessment
No full textBackground: Depression and falls are highly prevalent, interrelated concerns for nursing home (NH) residents. Relationships between depression and falls should guide nurses towards developing evidence-based practices for assessing these conditions together. This study aimed to ascertain NH nurses’ clinical practices and perceptions regarding the assessment of depression and fall risk. Methods: This study was an exploratory descriptive study on the reported practices and perceptions from NH nurses in the canton of Vaud, Western Switzerland. Statistical analyses included descriptive statistics, nonparametric tests and a content analysis of responses to open-ended questions. Results: The mean age of our 116 responding nurses was 44.6 years old (SD = 11.3), 99 were women and their mean work experience in NHs was 13.1 years (SD = 9.2). The reporting showed that 88.8% of nurses relied on mood observation for assessing depression and 88.8% relied on the history of falls to identify fall risk. Only 75.9% and 61.2% of nurses used validated scales to detect depression and fall risk, respectively. Additionally, 56.9% of participants considered depression to be a significant factor in fall risk. Conclusion: Validated tools to assess depression and fall risk in NHs should be used more widely. Health policies must support and enhance NH nurses’ training and skills
All clear ::Copilot and higher education
No full textArtificial Intelligence (AI) has long been touted as a panacea for producing knowledge. Fed with millions of texts and sources, the popularity of AI Generators (Gen AI) has skyrocketed as they seamlessly produce copious amounts of text in seconds. The user enters a prompt, and the output is practically simultaneous. AI tools have made their way into every corner of business including marketing and communications. The question revolves around the quality of the text that is generated. Should professionals use these Gen AI tools or avoid them like the plague? In the past ten years, Gen AI has become 'smarter,' using human feedback to improve with each new version. One such tool is Copilot, developed by Microsoft and an add-on to existing Microsoft 365 packages. Unlike its more popular Gen AI competitor, ChatGPT, Copilot is more effective at contextualizing content on a company level, i.e., business writing. Further, it is 'safer' than many Gen AI that share personal information in their responses to any user. Nonetheless, one area in which Gen AI has thrived is academia. During the COVID-19 pandemic, we were obliged to teach and learn remotely and engage with technology to do so. However, using technology for teaching and learning is different from using Gen AI, which offers new possibilities to create (for students and faculty members) and assess (for faculty) content. This study aims to test Copilot's capability to complete an academic writing task when given a specific prompt. The author has yet to see any study investigating Copilot's potential in academic writing tasks. This paper attempts to fill that gap. The project began with one prompt: Write a 500-word response to the following position: "This paper posits that robots cannot replace humans in the hospitality industry; thus, hoteliers should invest more in promoting the human touch." The response must include five academic journal articles used in-text and on a reference list in APA 7th edition format. This prompt was run 50 times in a row to test the veracity of the quality of the responses. Each response was saved in a file and the researcher took notes of observations of the output. To further these initial observations, the researcher ran all prompts through WordStat to identify common themes in the responses. At this stage, early findings show many similarities between the texts, suggesting that students in the same class and using the same prompt would potentially have issues with similarity checks or academic integrity. Further, the presentation is not in essay format; rather, the output looks like business documents that list (with bullet points) the main ideas. This is not the format that one would expect of an academic writing assignment. This could confirm that Copilot may be useful for the workplace, but, in its current state, is not effective in an educational setting. This project will help faculty better understand how Copilot could or could not be useful for pedagogical tasks. What should they be looking for? Are there specific phrases or patterns that could suggest the use of AI? The findings will also incite discussion and debate regarding the use of Gen AI in the classroom
Always look on the bright side? ::exploring the role of problem telling in entrepreneurial narratives to establish legitimacy
No full textIn the field of entrepreneurship, storytelling and legitimacy play a fundamental role. While previous studies have explored the various facets of narratives in entrepreneurial ventures, this study focuses on how entrepreneurs use problem-telling as a strategic communication tool to enhance the legitimacy of the venture. Using a qualitative approach, we aim to uncover the strategies used by entrepreneurs to create effective narratives about challenging events and problems they encounter. This study contributes to a comprehensive understanding of legitimacy building in entrepreneurship and sheds light on the dynamic nature of problem-telling under uncertainty, offering valuable insights for entrepreneurs seeking to engage stakeholders and foster future growth through the strategic use of narratives
Factors influencing cervical cancer re-screening in a semi-rural health district of Cameroon ::a cohort study
No full textBackground : Screening participation at recommended intervals is a crucial component of cervical cancer prevention effectiveness. However, little is known regarding the rate of re-screening in a Sub-Saharan context. This study aimed to estimate the re-screening rate of women in a semi-rural after an initial HPV-based screening and identify factors that influence adherence. Methods : This cohort study at the Annex Regional Hospital of Dschang enrolled women screened for cervical cancer over 5 years ago and due for re-screening. Women who initially tested HPV-positive (n = 132) and a random sample of HPV-negative women (n = 220) participated in a telephone survey between October 2021 and March 2022 to assess re-screening participation and reasons. Sociodemographic factors were collected, and associations with rescreening were evaluated. Results : A total of 352 participants aged under 50 years (mean age 37.4 years) were contacted, and 203 (58.0%) completed the survey. The proportion of women who complied with the screening recommendation was 34.0% (95% CI 27.5% − 40.5%), The weighted re-screening proportion was 28.4%. Age, marital status, education level, type of employment, and place of residence were not associated with the rate of re-screening. Main reported barriers to re-screening were lack of information (39.0%), forgetfulness (39.0%), and impression of being in good health (30.0%). Women who remembered the recommended screening interval were 2 to 3 times more likely to undergo re-screening (aOR (adjusted odds ratio) = 2.3 [1.2–4.4], p = 0.013). Human papilloma virus- positive status at the initial screening was also associated with the re-screening((aOR) (95% CI): 3.4 (1.8–6.5). Conclusion : Following an initial Human Papilloma Virus-based screening campaign in the West Region of Cameroon, one third of women adhered to re-screening within the recommended timeframe. Existing screening strategies would benefit from developing better information approaches to reinforce the importance of repeated cervical cancer screening
Design and modeling of PV-integrated double skin facades and application to retrofit buildings
No full textDouble Skin Façade (DSF) system comprises two glazing layers with a ventilated cavity. Integrating photovoltaic (PV) modules within the outer layer of DSFs offers an efficient method for electricity generation. Current tools for modeling and analyzing DSF systems are complex and resource-intensive, lacking the capability to evaluate the performance of innovative PV-DSF systems during the early design stage. This study develops a mathematical model to evaluate the electrical and thermal performance of PV-DSF systems, considering architectural design elements such as PV color and relative orientation. Based on an energy balance approach, the model is particularly suited for designing PV-DSF systems in heritage buildings, which often have color and relative orientation constraints. The model is applied to assess the performance of PV-DSF systems with conventional clear glass PV and colored front glass PV modules under the climatic conditions of Montreal, Canada. Results indicated that conventional clear glass PV module exhibit higher PV cell temperature than colored PV modules due to greater transmissivity, with peak temperature differences at noon of 5.5 °C, 6.2 °C, and 6.5 °C for orange, blue, and gray PV modules, respectively. On the contrary, the influence of PV's color front glass on room air temperature is non-significant. Furthermore, the optimal orientation for maximum energy yield is not always south-facing; it depends on the hourly distribution of the beam, diffuse solar irradiation, and ambient air temperature. For Montreal, west-facing DSFs produce more electrical and thermal energy on a summer design day because the hourly distribution of beam radiation is skewed towards afternoon hours