Kaikki aineistot
Lisää
Abstract This study adopts a sociocultural approach to examine the language socialisation of a lower age novice-newcomer recently arrived in a Finnish open day care centre. A mixed method approach combining ethnographic analysis and quantitative analysis has been used to analyse video-recordings. The results suggest that when educators recognise language novice-newcomer’s diverse needs and deploy thereby multimodal socialising strategies (i.e. verbal, gestural), the novice-newcomer’s language and interactive competences could ameliorate across time. Different from adults, the local children seem unconscious of the diverse needs of peers from diverse cultural and linguistic backgrounds, which sets out challenges for educators in balancing recognition of diversity with equality.
Microfibers are key components in construction of fiber-based materials, biomimetic materials, microsensors, and other fiber-based microstructures. Due to the scaling law, the adhesion forces such as van der Waals force or electrostatic force in the micro world play a more dominant role than the gravity, causing difficulties in precise pick-and-place of the micro objects. In this paper, we propose a capillary force-based pick-and-place method for handling microfibers. The method combines the typical robotic transportation technique and capillary gripping method to achieve fast and accurate pick-and-place of microfibers. We quantitatively analyzed the critical conditions for capillary pick-up and placement of microfibers and validated the technique experimentally. The theoretical analysis indicates that both pick-up and the placement of microfibers are largely dependent on the contact length on the fiber or the contact angle of the meniscus on the substrate. The experimental results show that the microfibers can be reliably picked up from and placed on the substrate of different materials including paper tissue, glass, silicon, stainless steel, copper with droplet volume of 0.1 nL and 0.3 nL. We further applied the method to the placement of the microfibers on super hydrophilic-super hydrophobic grooves and studied its placement speed and accuracy. We demonstrated that microfiber can be placed in such grooves in less than 0.1 seconds, with linear placement accuracy of 10μ m and angular placement accuracy of 0.5°. The proposed method is fast and simple, and it is especially suitable for handling fragile and flexible micro sized objects and construction of fiber-based materials.
Abstract Weakly supervised learning can help local feature methods to overcome the obstacle of acquiring a large-scale dataset with densely labeled correspondences. However, since weak supervision cannot distinguish the losses caused by the detection and description steps, directly conducting weakly supervised learning within a joint training describe-then-detect pipeline suffers limited performance. In this paper, we propose a decoupled training describe-then-detect pipeline tailored for weakly supervised local feature learning. Within our pipeline, the detection step is decoupled from the description step and postponed until discriminative and robust descriptors are learned. In addition, we introduce a line-to-window search strategy to explicitly use the camera pose information for better descriptor learning. Extensive experiments show that our method, namely PoSFeat (Camera Pose Supervised Feature), outperforms previous fully and weakly supervised methods and achieves state-of-the-art performance on a wide range of downstream task.
Abstract With the significant improvement of the intelligent capabilities of smart devices accompanied by the increasingly high requirements. Edge computing is regarded as an effective solution to achieve rapid response by deploying applications and tasks close to users. However, many studies only consider complete offloading, or offload tasks to edge servers in any proportion when designing the allocation strategies, ignoring the dependencies between subtasks. To deal with the dynamic environment, some learning-based task allocation methods generally adopt a centralized training way, which leads to the excessive network transmission resource consumption, especially in the smart grid scenario. To tackle the aforementioned challenges, we investigate the collaborative task allocation (CTA) problem by jointly considering the difference between the benefit of the tasks execution under a certain allocation strategy and when all tasks are executed locally. In this paper, the objective is to maximize the system gain, and we propose an attention-aided federated learning algorithm to deal with the CTA problem, named AteFL, by learning a shared model and extracting the system context for better representing the network information. The simulation results also show the superiority of the proposed AteFL algorithm.
While chronological age is the single biggest risk factor for cancer, it is less clear whether frailty, an age-related state of physiological decline, may also predict cancer incidence. We assessed the associations of frailty index (FI) and frailty phenotype (FP) scores with the incidence of any cancer and five common cancers (breast, prostate, lung, colorectal, melanoma) in 453,144 UK Biobank (UKB) and 36,888 Screening Across the Lifespan Twin study (SALT) participants, who aged 38–73 years and had no cancer diagnosis at baseline. During a median follow-up of 10.9 and 10.7 years, 53,049 (11.7%) and 4,362 (11.8%) incident cancers were documented in UKB and SALT, respectively. Using multivariable-adjusted Cox models, we found a higher risk of any cancer in frail vs. non-frail UKB participants, when defined by both FI (hazard ratio [HR] = 1.22; 95% confidence interval [CI] = 1.17–1.28) and FP (HR = 1.16; 95% CI = 1.11–1.21). The FI in SALT similarly predicted risk of any cancer (HR = 1.31; 95% CI = 1.15–1.49). Moreover, frailty was predictive of lung cancer in UKB, although this association was not observed in SALT. Adding frailty scores to models including age, sex, and traditional cancer risk factors resulted in little improvement in C-statistics for most cancers. In a within-twin-pair analysis in SALT, the association between FI and any cancer was attenuated within monozygotic but not dizygotic twins, indicating that it may partly be explained by genetic factors. Our findings suggest that frailty scores are associated with the incidence of any cancer and lung cancer, although their clinical utility for predicting cancers may be limited.
While chronological age is the single biggest risk factor for cancer, it is less clear whether frailty, an age-related state of physiological decline, may also predict cancer incidence. We assessed the associations of frailty index (FI) and frailty phenotype (FP) scores with the incidence of any cancer and five common cancers (breast, prostate, lung, colorectal, melanoma) in 453,144 UK Biobank (UKB) and 36,888 Screening Across the Lifespan Twin study (SALT) participants, who aged 38–73 years and had no cancer diagnosis at baseline. During a median follow-up of 10.9 and 10.7 years, 53,049 (11.7%) and 4,362 (11.8%) incident cancers were documented in UKB and SALT, respectively. Using multivariable-adjusted Cox models, we found a higher risk of any cancer in frail vs. non-frail UKB participants, when defined by both FI (hazard ratio [HR] = 1.22; 95% confidence interval [CI] = 1.17–1.28) and FP (HR = 1.16; 95% CI = 1.11–1.21). The FI in SALT similarly predicted risk of any cancer (HR = 1.31; 95% CI = 1.15–1.49). Moreover, frailty was predictive of lung cancer in UKB, although this association was not observed in SALT. Adding frailty scores to models including age, sex, and traditional cancer risk factors resulted in little improvement in C-statistics for most cancers. In a within-twin-pair analysis in SALT, the association between FI and any cancer was attenuated within monozygotic but not dizygotic twins, indicating that it may partly be explained by genetic factors. Our findings suggest that frailty scores are associated with the incidence of any cancer and lung cancer, although their clinical utility for predicting cancers may be limited.
Introduction: Although frailty is commonly considered as a syndrome of old individuals, recent studies show that it can affect younger adults, too. Whether and how frailty differs in younger adults compared to old is however unknown. To this end, we analyzed the prevalence, characteristics, and risk factors of early-life (aged <65) and late-life (aged ≥65) frailty. Methods: We analyzed individuals in the UK Biobank (N = 405,123) and Swedish Screening Across the Lifespan Twin (SALT; N = 43,641) study. Frailty index (FI) scores ≥0.21 were used to demarcate frailty. Characteristics of early-life versus late-life frailty were analyzed by collating the FI items (deficits) into domains and comparing the domain scores between younger and older frail individuals. Logistic regression was used to assess the risk factors of frailty. Results: The pooled prevalence rates of frailty were 10.3% (95% confidence interval [CI]: 2.7-32.7), 14.4% (95% CI: 4.5-37.2), 19.2% (95% CI: 2.5-68.5) in individuals aged ≤55, 55-64, 65-74, respectively. Younger frail adults (aged <65) had higher scores in immunological, mental wellbeing, and pain-related domains, whereas older frail adults (aged ≥65) had higher scores in cardiometabolic, cancer, musculoskeletal, and sensory-related domains. Higher age, female sex, smoking, lower alcohol consumption, lower education, obesity, overweight, low income, and maternal smoking were similarly associated with the risk of early-life and late-life frailty. Conclusion: Frailty is prevalent also in younger age groups (aged <65) but differs in some of its characteristics from the old. The risk factors of frailty are nevertheless largely similar for early-life and late-life frailty.
Abstract Neutral network (NN) and clustering are the two commonly used methods for speech separation based on supervised learning. Recently, deep clustering methods have shown promising performance. In our study, considering that the spectrum of the sound source has time correlation, and the spatial position of the sound source has short-term stability, we combine the spectral and spatial features for deep clustering. In this work, the logarithmic amplitude spectrum (LPS) and the interaural phase difference (IPD) function of each time frequency (TF) unit for the binaural speech signal are extracted as feature. Then, these features of consecutive frames construct feature map, which are regarded as the input to the Bi-directional long short-term memory (BiLSTM). The feature maps are converted to the high-dimensional vectors through BiLSTM, which are used to classify the time-frequency units by K-means clustering. The clustering index are combined with mixed speech signal to reconstruct the target speech signal. The simulation results show that the proposed algorithm has a significant improvement in speech separation and speech quality, since the spectral and spatial information are all utilized for clustering. Also, the method is more generalized in untrained conditions compared with traditional NN method e.g., deep neural network (DNN) and convolutional neural networks (CNN) based method.
Abstract Traditional separation methods have limited ability to handle the speech separation problem in high reverberant and low signal-to-noise ratio (SNR) environments, and thus achieve unsatisfactory results. In this study, a convolutional neural network with temporal convolution and residual network (TC-ResNet) is proposed to realize speech separation in a complex acoustic environment. A simplified steered-response power phase transform, denoted as GSRP-PHAT, is employed to reduce the computational cost. The extracted features are reshaped to a special tensor as the system inputs and implements temporal convolution, which not only enlarges the receptive field of the convolution layer but also significantly reduces the network computational cost. Residual blocks are used to combine multiresolution features and accelerate the training procedure. A modified ideal ratio mask is applied as the training target. Simulation results demonstrate that the proposed microphone array speech separation algorithm based on TC-ResNet achieves a better performance in terms of distortion ratio, source-to-interference ratio, and short-time objective intelligibility in low SNR and high reverberant environments, particularly in untrained situations. This indicates that the proposed method has generalization to untrained conditions.
Background: C-reactive protein (CRP) is a sensitive biomarker of inflammation with moderate heritability. The role of rare functional genetic variants in relation to serum CRP is understudied. We aimed to examine gene mutation burden of protein-altering (PA) and loss-of-function (LOF) variants in association with serum CRP, and to further explore the clinical relevance. Methods: We included 161,430 unrelated participants of European ancestry from the UK Biobank. Of the rare (minor allele frequency <0.1%) and functional variants, 1,776,249 PA and 266,226 LOF variants were identified. Gene-based burden tests, linear regressions, and logistic regressions were performed to identify the candidate mutations at the gene and variant levels, to estimate the potential interaction effect between the identified PA mutation and obesity, and to evaluate the relative risk of 16 CRP-associated diseases. Results: At the gene level, PA mutation burdens of the CRP (β = −0.685, p = 2.87e-28) and G6PC genes (β = 0.203, p = 1.50e-06) were associated with reduced and increased serum CRP concentration, respectively. At the variant level, seven PA alleles in the CRP gene decreased serum CRP, of which the per-allele effects were approximately three to seven times greater than that of a common variant in the same locus. The effects of obesity and central obesity on serum CRP concentration were smaller among the PA mutation carriers in the CRP (pinteraction = 0.008) and G6PC gene (pinteraction = 0.034) compared to the corresponding non-carriers. Conclusion: PA mutation burdens in the CRP and G6PC genes are strongly associated with decreased serum CRP concentrations. As serum CRP and obesity are important predictors of cardiovascular risks in clinics, our observations suggest taking rare genetic factors into consideration might improve the delivery of precision medicine.