Search

Background: Aging has been considered a risk determinant for chronic periodontitis. The aim of the present study is to investigate the age effect on healing after a comprehensive non-surgical treatment (CNSPT). Methods: In the study, the clinical data of CNSPT in 2012 to 2014 at the Department of Periodontics, National Taiwan University Hospital, was collected. The collected data was stratified according to initial clinical attachment level (cut-off value of 5 mm), tooth site (incisors/canines, premolars, and molars), and age (20-40, 40-60, and > 60 year-old). The probing depth (PD) reduction, gingival recession (GR), clinical attachment level (CAL) gain, and keratinized tissue (KT) gain were evaluated. Results: The treatment data of 204 patients were analyzed. In the 20-40 year-old group, PD reduction in all teeth, and CAL gain in teeth with initial CAL ≥5 mm, were significantly greater relative to the > 60 year-old group. In teeth with initial CAL ≥5 mm , the linear regression analysis also revealed a significantly negative correlation between age and PD reduction/ CAL gain. Regarding tooth type, PD reduction and CAL gain were significantly negatively correlated with age in incisors and canines. Conclusions: Both PD reduction and CAL gain were slightly greater in younger receiving CNSPT, specifically for incisors and canines with initial CAL ≥ 5 mm. &nbsp

Medical vision-language models enable co-learning and integrating features from medical imaging and clinical text. However, these models are not easy to train and the latent representation space can be complex. Here we propose a novel way for pre-training and regularising medical vision-language models. The proposed method, named Medical vision-language pre-training with Frozen language models and Latent spAce Geometry optimization (M-FLAG), leverages a frozen language model for training stability and efficiency and introduces a novel orthogonality loss to harmonize the latent space geometry. We demonstrate the potential of the pre-trained model on three downstream tasks: medical image classification, segmentation, and object detection. Extensive experiments across five public datasets demonstrate that M-FLAG significantly outperforms existing medical vision-language pre-training approaches and reduces the number of parameters by 78\%. Notably, M-FLAG achieves outstanding performance on the segmentation task while using only 1\% of the RSNA dataset, even outperforming ImageNet pre-trained models that have been fine-tuned using 100\% of the data., Comment: Accepted by MICCAI 2023

Electrocardiography (ECG) signals can be considered as multi-variable time-series. The state-of-the-art ECG data classification approaches, based on either feature engineering or deep learning techniques, treat separately spectral and time domains in machine learning systems. No spectral-time domain communication mechanism inside the classifier model can be found in current approaches, leading to difficulties in identifying complex ECG forms. In this paper, we proposed a novel deep learning model named Spectral Cross-domain neural network (SCDNN) with a new block called Soft-adaptive threshold spectral enhancement (SATSE), to simultaneously reveal the key information embedded in spectral and time domains inside the neural network. More precisely, the domain-cross information is captured by a general Convolutional neural network (CNN) backbone, and different information sources are merged by a self-adaptive mechanism to mine the connection between time and spectral domains. In SATSE, the knowledge from time and spectral domains is extracted via the Fast Fourier Transformation (FFT) with soft trainable thresholds in modified Sigmoid functions. The proposed SCDNN is tested with several classification tasks implemented on the public ECG databases \textit{PTB-XL} and \textit{MIT-BIH}. SCDNN outperforms the state-of-the-art approaches with a low computational cost regarding a variety of metrics in all classification tasks on both databases, by finding appropriate domains from the infinite spectral mapping. The convergence of the trainable thresholds in the spectral domain is also numerically investigated in this paper. The robust performance of SCDNN provides a new perspective to exploit knowledge across deep learning models from time and spectral domains. The repository can be found: https://github.com/DL-WG/SCDNN-TS

In a number of machine learning models, an input query is searched across the trained class vectors to find the closest feature class vector in cosine similarity metric. However, performing the cosine similarities between the vectors in Von-Neumann machines involves a large number of multiplications, Euclidean normalizations and division operations, thus incurring heavy hardware energy and latency overheads. Moreover, due to the memory wall problem that presents in the conventional architecture, frequent cosine similarity-based searches (CSSs) over the class vectors requires a lot of data movements, limiting the throughput and efficiency of the system. To overcome the aforementioned challenges, this paper introduces COSIME, an general in-memory associative memory (AM) engine based on the ferroelectric FET (FeFET) device for efficient CSS. By leveraging the one-transistor AND gate function of FeFET devices, current-based translinear analog circuit and winner-take-all (WTA) circuitry, COSIME can realize parallel in-memory CSS across all the entries in a memory block, and output the closest word to the input query in cosine similarity metric. Evaluation results at the array level suggest that the proposed COSIME design achieves 333X and 90.5X latency and energy improvements, respectively, and realizes better classification accuracy when compared with an AM design implementing approximated CSS. The proposed in-memory computing fabric is evaluated for an HDC problem, showcasing that COSIME can achieve on average 47.1X and 98.5X speedup and energy efficiency improvements compared with an GPU implementation., Comment: Accepted by the 41st International Conference on Computer Aided Design (ICCAD), San Diego, USA

Human Rhinoviruses (RVs) are dominant pathogens causing a wide range of respiratory tract diseases, posing a huge threat to public health worldwide. Viruses belonging to the RV-C species are more likely to cause severe illnesses and are strongly associated with asthma onset or exacerbations than RV-A or RV-B. Rapid and sensitive detection of neutralizing antibodies (NAbs) against RV-C can promote the development of vaccines and antiviral drugs and help in the diagnosis of viral infection. In this study, a rapid neutralization testing system for RV-C15, based on an enzyme-linked immunospot assay (Nt-ELISPOT) was developed. A monoclonal antibody (MAb), named 9F9, with high binding efficacy for RV-C15 conjugated to horseradish peroxidase (HRP), was used to detect RV-C15-infected cells at a concentration of 2 μg/ml. The optimal infectious dose of RV-C15 was set at 1 × 104 TCID50/well and the cells were fixed with 0.5% formaldehyde diluted in PBS after incubation for 20 h. Compared with the traditional cytopathic effect (CPE)-based neutralization assay (Nt-CPE), Nt-ELISPOT significantly shortened the detection period and showed good consistency with the detection of neutralizing titers of both sera and NAbs. Using Nt-ELISPOT, three anti-RV-C15 NAbs were obtained with IC50 values of 0.16, 0.27, and 11.8 μg/ml, respectively. Moreover, 64 human serum samples collected from a wide range of age groups were tested for NAb against RV-C15 by Nt-ELISPOT. The total seroprevalence was 48.4% (31/64) and the positive rate was lowest in the group under 6 years old. Thus, the Nt-ELISPOT established in this study can be used as a high-throughput and rapid neutralization assay for the screening of NAbs and for seroepidemiological investigation against RV-C15.

In this paper, we introduce the task of learning unsupervised dialogue embeddings. Trivial approaches such as combining pre-trained word or sentence embeddings and encoding through pre-trained language models (PLMs) have been shown to be feasible for this task. However, these approaches typically ignore the conversational interactions between interlocutors, resulting in poor performance. To address this issue, we proposed a self-guided contrastive learning approach named dial2vec. Dial2vec considers a dialogue as an information exchange process. It captures the conversational interaction patterns between interlocutors and leverages them to guide the learning of the embeddings corresponding to each interlocutor. The dialogue embedding is obtained by an aggregation of the embeddings from all interlocutors. To verify our approach, we establish a comprehensive benchmark consisting of six widely-used dialogue datasets. We consider three evaluation tasks: domain categorization, semantic relatedness, and dialogue retrieval. Dial2vec achieves on average 8.7, 9.0, and 13.8 points absolute improvements in terms of purity, Spearman's correlation, and mean average precision (MAP) over the strongest baseline on the three tasks respectively. Further analysis shows that dial2vec obtains informative and discriminative embeddings for both interlocutors under the guidance of the conversational interactions and achieves the best performance when aggregating them through the interlocutor-level pooling strategy. All codes and data are publicly available at https://github.com/AlibabaResearch/DAMO-ConvAI/tree/main/dial2vec., Comment: Accepted as Long Paper at "EMNLP,2022"

We study the performance of Hyper-Flex-LION optical interconnect architecture under dynamic traffic with traffic-prediction-aided multi-cluster reconfiguration. The simulation results show a 17.2% latency improvement and 36.9% packet loss reduction as compared to a fixed topology.

Learning sentence embeddings from dialogues has drawn increasing attention due to its low annotation cost and high domain adaptability. Conventional approaches employ the siamese-network for this task, which obtains the sentence embeddings through modeling the context-response semantic relevance by applying a feed-forward network on top of the sentence encoders. However, as the semantic textual similarity is commonly measured through the element-wise distance metrics (e.g. cosine and L2 distance), such architecture yields a large gap between training and evaluating. In this paper, we propose DialogueCSE, a dialogue-based contrastive learning approach to tackle this issue. DialogueCSE first introduces a novel matching-guided embedding (MGE) mechanism, which generates a context-aware embedding for each candidate response embedding (i.e. the context-free embedding) according to the guidance of the multi-turn context-response matching matrices. Then it pairs each context-aware embedding with its corresponding context-free embedding and finally minimizes the contrastive loss across all pairs. We evaluate our model on three multi-turn dialogue datasets: the Microsoft Dialogue Corpus, the Jing Dong Dialogue Corpus, and the E-commerce Dialogue Corpus. Evaluation results show that our approach significantly outperforms the baselines across all three datasets in terms of MAP and Spearman's correlation measures, demonstrating its effectiveness. Further quantitative experiments show that our approach achieves better performance when leveraging more dialogue context and remains robust when less training data is provided., Accepted as Long Paper at "EMNLP,2021"

Residential electricity consumption accounts for a large proportion of the primary energy consumption in China. Building energy management can effectively improve energy efficiency, where the accurate and reliable prediction results of residential electricity consumption are vital for the optimal management of building energy. In this paper, a hybrid model is developed for the ultra-short-term predictions of residential electricity consumption based on the Holt-Winters (HW) method and Extreme Learning Machine (ELM) network. The original data are decomposed into a stationary linear component and a fluctuant nonlinear residual using the Moving Average (MA) filter. The HW method is responsible for establishing the linear prediction model to forecasting the linear component. With the linear prediction results, nonlinear residual, and original data as inputs, the ELM could build a nonlinear prediction model for residential electricity consumption. The proposed HW-ELM model was used to predict 15-minute electricity consumption values in different training set sizes and seasons. When predicting residential electricity consumption, in comparison with HW, ELM, and long short-term memory network, the proposed model commonly demonstrated lower error. For a training set size of 50 days in spring, the root mean square error values were reduced by 87.98%, 64.89%, and 53.39%, respectively. The results of the experiments show that the proposed HW-ELM model offers more outstanding performance compared with the established models by the other methods.

Energy storage can address the mismatch of the ratio of heat to electricity between a combined cooling, heating, and power (CCHP) system and its users, and thus, it can significantly improve energy efficiency. However, energy storage also increases the complexity of the operation optimization of the system. Existing heuristic optimization algorithms such as genetic algorithm (GA) and particle swarm optimization can hardly obtain the optimal scheduling scheme. In this paper, a hybrid optimization method that combines the GA and dynamic programming (DP) is proposed. The GA is the main optimization framework and is used to optimize the hourly set points of the power generation unit in a day. In the optimization process, the GA generates a feasible solution set, and calls the DP to calculate the optimal energy storage set points for each solution. The DP defines an hour as a decision step, and enumerates all energy storage states in each decision step. This process loops until the optimal solution is obtained. To reduce the computing time, the DP is implemented as a vectorized code. Case studies are conducted to verify the effectiveness of the proposed method. The results demonstrate that the overall performance using the proposed method increases by 1.92% in summer and by 1.91% in winter compared with that using the traditional GA method. Furthermore, the computing time is acceptable for the scheduling of the energy system. The proposed method can also be applied to the operation optimization of the CCHP system considering the demand side response.

Summary Cell and organelle membranes consist of a complex mixture of phospholipids (PLs) that determine their size, shape, and function. Phosphatidylcholine (PC) is the most abundant phospholipid in eukaryotic membranes, yet how cells sense and regulate its levels in vivo remains unclear. Here we show that PCYT1A, the rate-limiting enzyme of PC synthesis, is intranuclear and re-locates to the nuclear membrane in response to the need for membrane PL synthesis in yeast, fly, and mammalian cells. By aligning imaging with lipidomic analysis and data-driven modeling, we demonstrate that yeast PCYT1A membrane association correlates with membrane stored curvature elastic stress estimates. Furthermore, this process occurs inside the nucleus, although nuclear localization signal mutants can compensate for the loss of endogenous PCYT1A in yeast and in fly photoreceptors. These data suggest an ancient mechanism by which nucleoplasmic PCYT1A senses surface PL packing defects on the inner nuclear membrane to control PC homeostasis., Graphical Abstract, Highlights • PCYT1A catalyzes the rate-limiting reaction in phosphatidylcholine (PC) synthesis • PC deficiency increases membrane lipid stored curvature elastic (SCE) stress • PCYT1A uses a nascent amphipathic helix to sense SCE in the nuclear membrane • PCYT1A mutants lacking a nuclear localization signal rescue PCYT1A-deficient cells, Phosphatidylcholine (PC) is the major constituent of cellular membranes. Haider et al. show that PCYT1A, the rate-limiting enzyme of PC synthesis, translocates onto the inner nuclear membrane in response to increased PC need in yeast, fly, and mammalian cells. This PCYT1A re-localization is governed by membrane stored curvature elastic stress.

Insulator strings is the important equipment in power system. It is necessary to identify the insulator string for completing the cleaning task of insulator string. A new method is proposed to identify insulator strings by multi sensor fusion algorithm based on D-S evidence theory. Firstly, the gray characteristic curve is defined in the optical image of camera, and the gray repetitive matrix and gray threshold matrix is constructed by the gray characteristic curve. Then, the depth characteristic curve is defined in the depth image of laser radar, and the depth repetitive matrix and width feature matrix is constructed by the depth characteristic curve. The features of every matrix are fussed by the Dempster-Shafer evidence theory to complete automatic recognition of three types isolators. The experimental results show that the new method can correctly identify the insulator strings under different lighting conditions.

In the era of Internet of Things (IoT) and with the explosive worldwide growth of electronic data volume, and associated need of processing, analysis, and storage of such humongous volume of data, several new challenges are faced in protect-ing privacy of sensitive data and securing systems by designing novel schemes for secure authentication, integrity protection, encryption, and non-repudiation. Lightweight symmetric key cryptography and adaptive network security algo-rithms are in demand for mitigating these challenges. This book presents some of the state-of-the-art research work in the field of cryptography and security in computing and communications. It is a valuable source of knowledge for re-searchers, engineers, practitioners, graduates, and doctoral students who are working in the field of cryptography, network security, and security and privacy issues in the Internet of Things (IoT). It will also be useful for faculty members of graduate schools and universities., Comment: 190 pages, 8 Chapters. Published by Intech Open Publishers, Croatia, July 2017

Under the background of information networking, the Marxism Theory Education is faced with challenges and opportunities. So the content, way, method and discipline system of Marxism Theory Education must be innovated. Under the background of information networking, Marxism Theory Education should draw on advantages and avoid disadvantages to adapt to new situation. Besides, it should make the education not only have the characteristics of times but also embody Chinese characteristics, hereby continuously create a new situation of Marxism Theory Education.

Precision measurement technology is the important measure of a nation's industrial technology level. However, with the rapid development of modern science and technology, metrological grating directly used for displacement measurement has could not meet the needs of contemporary people to the accuracy of measurement. So it is necessary to subdivide the grating sensor signal, especially the fine subdivision which has been increasingly concerned. Aiming at the constructor method used in the specific implementation process of the fine subdivision of the grating sensor signal, two schemes are put forward in this paper. Based on the analysis and comparison of the two methods, at last, the optimal design scheme will be determined.

The empirically successful Thompson Sampling algorithm for stochastic bandits has drawn much interest in understanding its theoretical properties. One important benefit of the algorithm is that it allows domain knowledge to be conveniently encoded as a prior distribution to balance exploration and exploitation more effectively. While it is generally believed that the algorithm's regret is low (high) when the prior is good (bad), little is known about the exact dependence. In this paper, we fully characterize the algorithm's worst-case dependence of regret on the choice of prior, focusing on a special yet representative case. These results also provide insights into the general sensitivity of the algorithm to the choice of priors. In particular, with $p$ being the prior probability mass of the true reward-generating model, we prove $O(\sqrt{T/p})$ and $O(\sqrt{(1-p)T})$ regret upper bounds for the bad- and good-prior cases, respectively, as well as \emph{matching} lower bounds. Our proofs rely on the discovery of a fundamental property of Thompson Sampling and make heavy use of martingale theory, both of which appear novel in the literature, to the best of our knowledge., Comment: Appears in the 27th International Conference on Algorithmic Learning Theory (ALT), 2016

With the tendency of population aging continues in our nation, problems associated with the elderly become major concerns in our society. If daycare centers for seniors can be established to care for the elderly just like daycare facilities for caring pre-school age children, seniors will have a place to enjoy day-time activities and return home to enjoy normal family life. The senior center that allows seniors to enjoyable their remaining years with dignity will be a demanding neighborhood facility. However, the location of such a facility is of great importance for its acceptance by the general public; a location that pleases seniors and their family is expected by all involved. The senior daycare center is a fixed and permanent facility; its location is also the key to affect the return of investors. A properly selected location will maximize the economic benefit while lowering the risks associated with the investment. Hence, in this research, the CFA (Confirmatory Factor Analysis) available in SEM (Structural Equation Modeling) is applied to evaluate important factors that affect the selection of senior daycare center and the degree of suitability for implementing the model to conduct such evaluations. The results on selecting the most appropriate location for establishing a senior daycare center are verified with case study data using the extension method. The combination of structural equation modeling and extension method leads to results that comfort to model evaluation indices when the suitability of individual models and the overall modeling are concerned. Using the extension method allows one to find the most appropriate location based on case study information. This proves that the method implemented in this research is capable of solving problems that are contradictory and incompatible in nature.

Description

Description

Phototransduction in Drosophila is mediated by phospholipase C-dependent hydrolysis of PIP2-, and is an important model for phosphoinositide signalling. Although generally assumed to operate by generic machinery conserved from yeast to mammals, some key elements of the phosphoinositide cycle have yet to be identified in Drosophila photoreceptors. Here, we used transgenic flies expressing fluorescently tagged probes (P4M and TbR332H), which allow in vivo quantitative measurements of PI4P and PIP2 dynamics in photoreceptors of intact living flies. Using mutants and RNA interference for candidate genes potentially involved in phosphoinositide turnover, we identified Drosophila PI4KIIIα (CG10260) as the PI4-kinase responsible for PI4P synthesis in the photoreceptor membrane. Our results also indicate that PI4KIIIα activity requires rbo (the Drosophila orthologue of Efr3) and CG8325 (orthologue of YPP1), both of which are implicated as scaffolding proteins necessary for PI4KIIIα activity in yeast and mammals. However, our evidence indicates that the recently reported central role of dPIP5K59B (CG3682) in PIP2 synthesis in the rhabdomeres should be re-evaluated; although PIP2 resynthesis was suppressed by RNAi directed against dPIP5K59B, little or no defect was detected in a reportedly null mutant (dPIP5K18 ).

Objective: More than 300 genetic variants have been robustly associated with measures of human adiposity. Highly penetrant mutations causing human obesity do so largely by disrupting satiety pathways in the brain and increasing food intake. Most of the common obesity-predisposing variants are in, or near, genes that are expressed highly in the brain, but little is known about their function. Exploring the biology of these genes at scale in mammalian systems is challenging. We sought to establish and validate the use of a multicomponent screen for feeding behaviour phenotypes taking advantage of the tractable model organism Drosophila melanogaster. Methods: We validated a screen for feeding behaviour in Drosophila by comparing results after disrupting expression of centrally-expressed genes that influence energy balance in flies to those of ten control genes. We then used this screen to explore the effects of disrupted expression of genes that are either a) implicated in energy homeostasis through human genome wide association studies (GWAS) or b) expressed and nutritionally responsive in specific populations of hypothalamic neurons with a known role in feeding/fasting. Results: Using data from the validation study to classify responses, we studied 53 Drosophila orthologues of genes implicated by human GWAS in body mass index (BMI) and found that 15 significantly influenced feeding behaviour or energy homeostasis in the Drosophila screen. We then studied 50 Drosophila homologues of 47 murine genes that are reciprocally nutritionally regulated in POMC and AgRP neurons. Seven of these 50 genes were found by our screen to influence feeding behaviour in flies. Conclusion: Here, we demonstrate the utility of Drosophila as a tractable model organism in a high-throughput genetic screen for food intake phenotypes. This simple and cost-efficient strategy is ideal for high-throughput interrogation of genes implicated in feeding behaviour and obesity in mammals, and will facilitate the process of reaching a functional understanding of obesity pathogenesis.