Non-IID data present a tough challenge for federated learning. In this paper, we explore novel idea of facilitating pairwise collaborations between clients with similar data. We propose FedAMP, new method employing attentive message passing to facilitate collaborate more. establish the convergence FedAMP both convex and non-convex models, heuristic further improve performance when adopt deep neural networks as personalized models. Our extensive experiments on benchmark sets demonstrate...

The state-of-the-art cloud computing platforms are facing challenges, such as the high volume of crowdsourced data traffic and highly computational demands, involved in typical deep learning applications. More recently, Edge Computing has been recently proposed an effective way to reduce resource consumption. In this paper, we propose edge framework by introducing concept demonstrate superiority our on reducing network running time.

Due to the ever-growing demands in modern cities, unreliable and inefficient power transportation becomes one critical issue nowadays grid. This makes grid monitoring of key modules system play an important role preventing severe safety accidents. However, traditional manual inspection cannot efficiently achieve this goal due its low efficiency high cost. Smart as a new generation grid, sheds light construct intelligent, reliable efficient with advanced information technology. In smart...

Hepatitis B surface antigen (HBsAg) seroclearance during treatment is associated with a better prognosis among patients chronic hepatitis (CHB). Significant gaps remain in our understanding on how to predict HBsAg accurately and efficiently based obtainable clinical information. This study aimed identify the optimal model seroclearance. We obtained laboratory demographic information for 2,235 CHB from South China Monitoring Administration (SCHEMA) cohort. occurred 106 total. developed models...

Non-IID data present a tough challenge for federated learning. In this paper, we explore novel idea of facilitating pairwise collaborations between clients with similar data. We propose FedAMP, new method employing attentive message passing to facilitate collaborate more. establish the convergence FedAMP both convex and non-convex models, heuristic further improve performance when adopt deep neural networks as personalized models. Our extensive experiments on benchmark sets demonstrate...

With the deep penetration of mobile devices, more and learning applications have been widely used in daily life. However, since tasks are computationally intensive, limited computation resource on devices cannot execute application effectively. The common approaches transmitting data from offloading to cloud. This brings another issue that high transmission delay may become bottleneck performance. In this paper, we explore a new rising concept, edge computing, into applications. Comparing...

Deep learning has been applied in many recent advanced applications the field of transportation, finance and medicine. These require significant computation resources large-scale training samples. Cloud becomes a natural choice for conducting these tasks due to its abundant resources. However, deeper penetration deep techniques mission critical applications, like driverless car, calls stricter time requirement guarantee interaction larger amount dataset accuracy, which cannot be easily...

Abstract Semi-quantum key distribution describes a system in which fully quantum user and classical perform distribution. The main advantage of is its security. Owing to the bottlenecks existing technology, highly attenuated lasers threshold detectors are required for semi-quantum distribution; however, these components make susceptible eavesdroppers. Our previous study presented first experiment verified feasibility mirror protocol 2021. Herein, we build channel model use...

Abstract Semi-quantum key distribution (SQKD) is used to establish a string of shared secret keys between quantum party and classical party. Here, we report the first proof-of-principle experimental demonstration SQKD based on Mirror protocol, which most experimentally feasible equipped with time-phase encoding scheme employing method selective modulation. The experiment was performed at repetition frequency 62.5 MHz high raw rate arrived 69.8 kbps, average bit error found be 4.56% 2.78% for...

Intrinsic circadian clocks generate rhythms of physiology and behavior, which provide the capabilities to adapt cycling environmental cues that result from self-rotation Earth. Circadian misalignment leads deleterious impacts on adaptation health in different organisms. The interplanetary journey Mars dramatically differ those These differences impose numerous adaptive challenges, including challenges for humans’ clock. Thus, Martian environment is a prerequisite future landing dwelling...

Over the past years, Human Activity Recognition (HAR) has shown its great value and been further developed with help of deep learning. However, existing HAR systems that use learning methods to achieve ideal accuracy recognition heavily rely on massive amounts labeled training samples. Unfortunately, it requires considerable human effort is unrealistic for real-life applications. In this paper, we propose a novel system, which combines active WiFi-based HAR. The system capable building good...

Since 2019, COVID-19 began spreading globally and has significantly affected peoples’ daily lifestyles. The public was asked to stay at home for constant quarantine community containment starting on 23 January 2020. To assess the circadian rhythms sleep changes their influential factors during outbreak, a questionnaire administered 451 Chinese participants 20–31 in rhythm, sleep–wake cycle, dining, exercise of correlation with negative emotions were analyzed. Furthermore, effects three...

Recently, short texts become very popular in social life. To understand texts, researchers develop topic models to extract information. However, conventional mainly focus on long documents which cannot deal with the sparsity problem of text. In this paper, we propose a novel model for text called GPU-BTM, incorporates Generalized Pólya Urn technique into Biterm Topic Model. GPU-BTM utilizes similarity information and co-occurrence pattern words simultaneously handle problem. Specifically,...

In this society, the way people obtain information has gradually shifted from traditional media to social applications. Social applications publish a large number of real-world events every day. With more and about in networks, is contained. event detection refers extracting interrelated message clusters corpus or stream represent specific real world. Combining with different domain knowledge can research mine information. article, we will explore based on topic models, community discovery,...

Background: Lung cancer is the most aggressive which representing one-quarter of all cancer-related deaths, and metastatic spread accounts for >70% these especially brain metastasis. Metastasis associated mutations are important biomarkers metastasis prediction outcome improvement. Methods: In this study, we applied whole-exome sequencing to identify potential related mutation in 12 paired lung samples. Findings: We identified 1,702 SNVs 6,131 events 1,220 genes. Furthermore, several...

Short text topic modeling attracts many researchers' attention with the emergence of online social media platforms, such as news websites, Twitter and Facebook. Existing models for short texts mainly focus on relieving sparse problem to enhance accuracy performance modeling. However, most previous approaches introduce external corpus word embeddings enrich global semantic information in process, ignoring local association target corpus. And provided by embedding may not be entirely suitable...

Events spreading on social media platforms reflect current public concerns and emotions among opinions. Heterogeneous elements of networks the sparse context messages bring significant challenges to fine-grained event detection task. Few existing methods can learn inherent structure rich semantics messages, nor they effectively update model in a dynamic scenario for continuously coming messages. In this paper, we design novel Multi-Semantics Graph Neural Network (MSGNN) events continuous...