Description:
Detecting protein complexes from protein-protein interaction (PPI) networks allows biologists reveal the principle of cellular organization and functions. Existing computational methods try to incorporate biological evidence to enhance the quality of predicted complexes. However, it is still a challenge to integrate biological information into complexes discovery process under a unified framework. Recently, network embedding methods showed their effectiveness in graph data analysis tasks. It provides a framework for incorporating both network structure and additional node attribute information. This salient feature is particularly desirable in the context of protein complexes identification. However, none of the existing network embedding methods take node attribute proximity and high-order structure proximity into account at the same time. In this paper, we propose a novel global network embedding method, which preserves global network structure and biological information. We utilize this global representation learning method to learn vector representation for proteins. Then, we use a seed-extension clustering method to discover overlapping protein complexes with the embedding results. This novel protein complexes detection method we called GLONE. Evaluated on five real yeast PPI networks, our method outperforms the competing algorithms in terms of different evaluation metrics.
Description:
Deep learning on graphs has attracted significant interests recently. However, most of the works have focused on (semi-) supervised learning, resulting in shortcomings including heavy label reliance, poor generalization, and weak robustness. To address these issues, self-supervised learning (SSL), which extracts informative knowledge through well-designed pretext tasks without relying on manual labels, has become a promising and trending learning paradigm for graph data. Different from SSL on other domains like computer vision and natural language processing, SSL on graphs has an exclusive background, design ideas, and taxonomies. Under the umbrella of graph self-supervised learning, we present a timely and comprehensive review of the existing approaches which employ SSL techniques for graph data. We construct a unified framework that mathematically formalizes the paradigm of graph SSL. According to the objectives of pretext tasks, we divide these approaches into four categories: generation-based, auxiliary property-based, contrast-based, and hybrid approaches. We further describe the applications of graph SSL across various research fields and summarize the commonly used datasets, evaluation benchmark, performance comparison and open-source codes of graph SSL. Finally, we discuss the remaining challenges and potential future directions in this research field. IEEE
Description:
Deep learning on graphs has attracted significant interests recently. However, most of the works have focused on (semi-) supervised learning, resulting in shortcomings including heavy label reliance, poor generalization, and weak robustness. To address these issues, self-supervised learning (SSL), which extracts informative knowledge through well-designed pretext tasks without relying on manual labels, has become a promising and trending learning paradigm for graph data. Different from SSL on other domains like computer vision and natural language processing, SSL on graphs has an exclusive background, design ideas, and taxonomies. Under the umbrella of graph self-supervised learning, we present a timely and comprehensive review of the existing approaches which employ SSL techniques for graph data. We construct a unified framework that mathematically formalizes the paradigm of graph SSL. According to the objectives of pretext tasks, we divide these approaches into four categories: generation-based, auxiliary property-based, contrast-based, and hybrid approaches. We further describe the applications of graph SSL across various research fields and summarize the commonly used datasets, evaluation benchmark, performance comparison and open-source codes of graph SSL. Finally, we discuss the remaining challenges and potential future directions in this research field. IEEE
Description:
The amount of time between key points in the work disability and workers' compensation process, referred to as lag times, has been shown to relate to work disability outcomes but little research has examined how this finding may differ based on the diagnosis associated with the cause of work disability. To examine whether relationships between lag times in the work disability process and disability duration vary by diagnosis in a sample of workers' compensation claims. Using workers' compensation claims, Analysis of Covariance was used to estimate differences in disability duration across three lag times (days from the date of injury to: reporting the injury, seeking medical care, and starting lost work time paid by the workers' compensation insurer) and injury diagnosis groups (Work-related Musculoskeletal Disorders (WRMSD) and fractures). WRMSD tended to have longer lag times than fractures, whereas disability duration tended to be longer for fractures than WRMSD. Overall, shorter lag times were associated with shorter disability duration, but the relationships varied across diagnosis groups, and greater variation in disability duration was observed for WRMSD compared to fractures. The findings suggest the importance of responding to work-related injuries by reporting the injury, receiving medical care, and taking time off of work if necessary, in a timely fashion.
Description:
This paper aims to develop a real-time Plant Environment Simulator (PES), which simulates a corrugated plant effectively and realistically. The resultant solution of this work can be used to provide factory workers or new developers with a responsive, simulated learning environment on teaching how to use existing software correctly. The work is carried out for a large cardbox maker that can be used to test new prototypes without using the actual plant facilities, so it will economically and efficiently contribute to the creation of new robust software products for the corrugated plant.
Description:
To address the upcoming reform curriculum initiative following competency-based learning in Vietnam, this study examines how variations of a contextualized task allow students to experience the mathematical modelling process. We explore how Grade 10 students respond to tasks with different levels of authenticity when they first attempt such tasks. Adapting the Theory of Authentic Task Situations, we designed variations of the Goat Cage task and implemented them with six small groups of students. Results showed that the modelling processes that the students progressed through differed in mathematical models used and real-world considerations, with diverse approaches when solving the more-authentic variation. Not all phases of the modelling process were observed explicitly when they engaged in the tasks. Students showed a positive disposition towards tasks with more authentic contextual information. The study suggests ways to adapt contextualized tasks to make them more authentic, which allows students to experience a holistic modelling process.