tulokset
Silmäile
Osuva
Osuva on Vaasan yliopiston avoin julkaisuarkisto. Osuva sisältää Vaasan yliopiston omat julkaisut, opinnäytteet ja tieteellisten artikkeleiden rinnakkaistallenteet. Osuvaan sisältyy julkaisujen viitetietoja, tiivistelmiä ja kokotekstejä. Sähköisten arkistokokoelmien sisältö ei ole luettavissa verkossa.
Kokoelmat Osuvassa
Viimeksi tallennetut
IBAC-Net : integrative brightness adaptive plant leaf disease classification
Xu, Xing; Ma , Hongya; Zhao , Yun; Lv, Xiaoshu (Pagepress, 2025-03-11)
As agricultural technology continues to advance, effective classification of agricultural diseases are crucial for improving crop yield and quality. This study aims to explore an innovative approach to agricultural disease image classification based on a novel image classification model architecture. First, we design a novel model architecture for image classification that better integrates shallow and deep features. Secondly, to address potential brightness differences in images collected under varying weather conditions, we have introduced an image brightness adaptive block. This block automatically adjusts the brightness of images during the data collection and processing stages, thereby reducing image disparities caused by weather variations. This step is crucial for improving the robustness of the model and ensuring accurate identification of agricultural diseases under different environmental conditions. Additionally, drawing inspiration from the Inception architecture and employing a flexible downsampling strategy, we have designed a custom inception block to integrate shallow and deep features effectively. To validate the effectiveness of our proposed approach, we conducted experiments using an agricultural disease image dataset processed with weather effects. The experimental results demonstrate that our model exhibits higher accuracy and robustness in agricultural disease image classification tasks compared to traditional methods. The code has been uploaded to GitHub at the following address: https://github.com/bettyaya/IBAC-Net.
Resilience of GVC suppliers in politically unstable regions : The roles of governance and trust
Choksy, Umair Shafi; Kurt, Yusuf; Gölgeci, Ismail; Khan , Zaheer; Shamim, Saqib; Jawad, Maaha (Elsevier, 2025-06-06)
Supplier firms in the Global South face compounded risks from political instability that challenge their ability to maintain participation in global value chains (GVCs). While resilience is increasingly acknowledged as a critical capability, it remains unclear how suppliers develop resilience when conventional GVC governance strategies, often grounded in institutional stability, prove insufficient. This gap is especially pronounced in knowledge-intensive service sectors like software development, where codification, coordination, and inter-firm trust are central but often disrupted by political instability. This study examines the resilience of GVC suppliers operating in politically unstable regions of South Asia, particularly India and Pakistan. It focuses on the role of governance mechanisms—such as codification, managing complexity, and supplier capabilities, on supplier resilience. It also investigates how trust moderates these relationships. Drawing on Transaction Cost Economics (TCE) and the Dynamic Capabilities View (DCV), we argue that these governance mechanisms function not only as efficiency enablers but also as dynamic governance adaptations that suppliers actively mobilize to survive and adapt. A quantitative analysis using partial least squares structural equation modeling (PLS-SEM) was conducted on survey data collected from 100 software firms. The results show that task codification and management of task complexity enhance supplier resilience. It further reveals that trust negatively affects the links between task codification and resilience, challenging the conventional assumption that trust uniformly strengthens GVC relationships. The core theoretical contribution of this study lies in extending TCE and DCV by showing how resilience is enabled through external governance adaptation and by rethinking trust as a conditional, context-dependent mechanism rather than a universal good.
Personalized Interactive Music Systems for Physical Activity and Exercise : Exploratory Systematic Review and Meta-Analysis
Danso, Andrew; Kekäläinen, Tiia; Koehler, Friederike; Knittle, Keegan; Nijhuis, Patti; Burunat, Iballa; Neto, Pedro; Mavrolampados, Anastasios; Randall, William M.; Hansen, Niels Chr; Ansani, Alessandro; Rantalainen, Timo; Alluri, Vinoo; Hartmann, Martin; Schaefer, Rebecca S.; Ihalainen, Johanna K.; Rousi, Rebekah; Agres, Kat R.; MacRitchie, Jennifer; Toiviainen, Petri; Saarikallio, Suvi; Chastin, Sebastien; Luck, Geoff (JMIR Publications, 2025-09-08)
Background:
Personalized Interactive Music Systems (PIMSs) are emerging as promising devices for enhancing physical activity and exercise outcomes. By leveraging real-time data and adaptive technologies, PIMSs align musical features, such as tempo and genre, with users’ physical activity patterns, including frequency and intensity, enhancing their overall experience.
Objective:
This exploratory systematic review and meta-analysis evaluates the effectiveness of PIMSs across physical, psychophysical, and affective domains.
Methods:
Searches across 9 databases identified 18 eligible studies, of which 6 (comprising 17 intervention arms) contained sufficient data for meta-analysis. Random-effects meta-analyses and meta-regression were performed to assess outcomes for physical activity levels, physical exertion, ratings of perceived exertion, and affective valence.
Results:
Results showed significant improvements in physical activity levels (g=0.49, CI 0.07 to 0.91, P=.02, k=4) and affective valence (g=1.65, CI 0.35 to 2.96, P=.01, k=4), with faster music tempo identified as a significant moderator (P=.03). No significant effects were observed for ratings of perceived exertion (g=0.72, CI −0.13 to 1.58, P=.10, k=3) or physical exertion (g=0.78, CI −0.55 to 2.11, P=.25, k=5).
Conclusions:
Substantial heterogeneity and limited study quality indicate the need for more robust, randomized controlled trials to establish the efficacy of PIMSs in diverse populations.
On the Estimation of Optimal Cutoffs for Power Laws and the Cross Section of Realized Foreign Exchange Rate Variances
Grobys, Klaus (Springer, 2025-09-05)
Extending recent research, this study introduces a novel testing procedure based on modern block bootstrap techniques and maximum likelihood estimation to investigate whether the universal power-law process governing the cross-section of realized foreign exchange (FX) rate variances exhibits a conjoint cutoff. The analysis posits that the maximum likelihood estimator for the exponent of a power law is intrinsically dependent on the selected cutoff. Our innovative test, calibrated to the cross-section of realized daily FX variances, provides evidence for the existence of such a universal cutoff. The findings have significant implications for FX risk management. Specifically, they indicate that (a) the benefits of FX risk diversification may be more constrained than previously assumed, and (b) the extent of power-law behavior in the realized variance risk of the FX market may be substantially underestimated when conventional single-equation models are employed to determine the optimal cutoff for a power law.
A stablecoin that’s actually stable : A portfolio optimization approach
Grobys, Klaus; Junttila, Juha-Pekka; Kolari, James W. (Elsevier, 2025-08-28)
Stablecoins seek to address the high price fluctuations of unbacked cryptocurrencies, such as Bitcoin and Ether. However, recent studies as well as the collapse of stablecoin USTC (Terra) cast doubt on the stability of stablecoins. Using well-known Markowitz portfolio optimization methods, we combine five leading stablecoins into a global minimum variance portfolio that represents a stable aggregate stablecoin (SAS). We find that SAS is much more stable than its constituent stablecoins. Also, in a stress test adding USTC to the portfolio, SAS remains stable with a narrow price range over time. Importantly, the construction of SAS using modern diversification methods has practical implications for the ongoing development of central bank digital currencies (CBDCs).