Pre-defined Database Research Articles

Effect of prophylactic corticosteroids on postoperative neurocognitive dysfunction in the adult population: An updated systematic review, meta-analysis, and trial sequential analysis of randomised controlled trials.

Postoperative neurocognitive dysfunction (PNCD) commonly occurs after surgery and prolongs hospital stays. Both direct noxious stimuli to the central nervous system and systemic inflammation have been implicated. Due to their potent anti-inflammatory effects, corticosteroids have been utilised to attenuate the incidence and severity of PNCD. This systematic review and meta-analysis strived to evaluate the prophylactic role of perioperative corticosteroids for PNCD. A search was run in pre-defined databases for randomised controlled trials (RCTs) assessing the role of corticosteroids in preventing PNCD. The incidence of PNCD within 1 month was the primary outcome. Secondary outcomes included the use of antipsychotic medications for the treatment, postoperative infection, and hospital length of stay. The results are exhibited as odds ratio (OR) and the mean difference (MD) with 95% confidence interval (CI). Fifteen RCTs comprising 15,398 patients were included. The incidence of PNCD was significantly lower in the corticosteroid group than in the control group, with a pooled OR of 0.75 (95% CI 0.58, 0.96; P = 0.02; I2 = 66%). Trial sequential analysis showed the clinical benefit of corticosteroids in preventing PNCD; however, the requisite information size is still inadequate. The sub-group analysis supported the prophylactic effect of corticosteroids on delirium prevention but not on delayed neurocognitive recovery. Our meta-analysis revealed statistically significant protective effects of corticosteroids on the incidence of PNCD. However, further studies are still needed to confirm the protective role of this commonly used and relatively safe strategy for preventing PNCD.

External knowledge document retrieval strategy based on intention-guided and meta-learning for task-oriented dialogues

The external knowledge augmented task-oriented dialogues system is designed to cover the user requests beyond the predefined data base (DB) and application-specific interface (APIs). However, the existing methods mainly focus to improve the retrieval accuracy of the external knowledge but ignore the modeling of the entity-aware dialogue intention, which can’t fast locate the document-level external API in the practical scenarios. Therefore, an external knowledge document retrieval strategy based on the intention-guided and meta-learning for the task-oriented dialogue has been proposed in this paper. First, the entity-aware dialogue intention information is introduced and then a two-stage training method for the intention-guided knowledge document retrieval model is presented, which is able to effectively enhance the retrieval accuracy of the Bi-encoder retrieval model in the task-based dialogue system. Moreover, to further improve the generalization ability of the retrieval model based on the intention-guided, a meta-task based on the domain offset is constructed and make the Bi-encoder retrieval model get the more robust semantic representation through the meta-learning, which can improve the generalization capability of the proposed model in the invisible domains. Experimental results tested on the augmented MultiWOZ 2.1 dataset are shown to prove the correctness and effectiveness of the proposed strategy.

A KEYLESS ANTI-THEFT SECURITY SYSTEM FOR AUTOMOBILES USING OPENCV PYTHON

Automobile theft is on the rise all over the world these days. As a result, car owners are turning to an approach known as anti-theft security system to prevent thieves from taking their vehicles. As a result, this system is concerned with the design and implementation of a security system that uses GSM technology to prevent automotive theft. We will demonstrate a paradigm of a real-time anti-theft security system that can be easily constructed by car owners all around the world in this project. A GSM service and a microcontroller are used in this scenario. The password is input as a security option, and if it is approved, the door is unlocked. Face recognition is an instant security feature that analyses the drivers face to a pre-defined database and starts the ignition procedure if they match. In the event that the foregoing process fails, a smart SMS will be sent to the authorised individual.

Hybrid Online Model for Predicting Diabetes Mellitus

Modern healthcare systems have become smart by synergizing the potentials of wireless sensors, the medical Internet of things, and big data science to provide better patient care while decreasing medical expenses. Large healthcare organizations generate and accumulate an incredible volume of data continuously. The already daunting volume of medical information has a massive amount of diagnostic features and logged details of patients for certain diseases such as diabetes. Diabetes mellitus has emerged as along-haul fatal disease across the globe and particularly in developing countries. Exact and early diagnosis of diabetes from big medical data is vital for the deterrence of disease and the selection of proper therapy. Traditional machine learning-based diagnosis systems have been initially established as offline (non-incremental) approaches that are trained with a pre-defined database before they can be applied to handle prediction problems. The major objective of the proposed work is to predict and classify diabetes mellitus by implementing a Hybrid Online Model for Early Detection of diabetes disease (HOMED) using machine learning algorithms. Our proposed online (incremental) diabetes diagnosis system exploits (i) an Adaptive Principal Component Analysis (APCA) technique for missing value imputation, data clustering, and feature selection; and (ii) an enhanced incremental support vector machine (ISVM) for classification. The efficiency of HOMED is estimated on different performance metrics such as accuracy, precision, specificity, sensitivity, positive predictive value, and negative predictive value. Experimental results on Pima Indian diabetes dataset (768 samples: 500 non diabetic and 268 diabetic patients) reveal that HOMED considerably increases the classification accuracy and decreases computational complexity with respect to the offline models. The proposed system can assist healthcare professionals as a decision support system.

Context and knowledge aware conversational model and system combination for grounded response generation

End-to-end neural-based dialogue systems can potentially generate tailored and coherent responses for user inputs. However, most of existing systems produce universal and non-informative responses, and they have not gone beyond chitchat yet. To tackle these problems, 7th Dialog System Technology Challenges (DSTC7-Track2) was developed to focus on building a dialogue system that produces informational responses that are grounded on external knowledge. In this study, we propose a Memory-augmented Hierarchical Recurrent Encoder-Decoder, called MHRED, that grounded on both multi-turn dialogue context and external knowledge. Furthermore, we apply a combination of multiple dialogue systems. Our final system is an ensemble that combines three modules: a generation-based module, a retrieval-based module, and a reranking module. First, responses are generated by MHRED, and retrieved from a pre-defined database focusing on informativeness. Next, the reranking module sorts these candidates using several hand-crafted features, and finally it selects a response with the highest score. Therefore, this system can return diverse and meaningful responses from various perspectives. Experimental results show that our proposed MHRED outperforms strong baseline models and combining multiple dialogue systems significantly improves the automatic evaluation and human evaluations.

MetaSearch: Incremental Product Search via Deep Meta-Learning

With the advancement of image processing and computer vision technology, content-based product search is applied in a wide variety of common tasks, such as online shopping, automatic checkout systems, and intelligent logistics. Given a product image as a query, existing product search systems mainly perform the retrieval process using predefined databases with fixed product categories. However, real-world applications often require inserting new categories or updating existing products in the product database. When using existing product search methods, the image feature extraction models must be retrained and database indexes must be rebuilt to accommodate the updated data, and these operations incur high costs for data annotation and training time. To this end, we propose a few-shot incremental product search framework with meta-learning, which requires very few annotated images and has a reasonable training time. In particular, our framework contains a multipooling-based product feature extractor that learns a discriminative representation for each product, and we also design a meta-learning-based feature adapter to guarantee the robustness of the few-shot features. Furthermore, when expanding new categories in batches during a product search, we reconstruct the few-shot features by using an incremental weight combiner to accommodate the incremental search task. Through extensive experiments, we demonstrate that the proposed framework achieves excellent performance for new products while still guaranteeing the high search accuracy of the base categories after gradually expanding new product categories without forgetting.

Big Data and Official Statistics

Big data is a component of the Fourth Industrial Revolution. The deep penetration of digital technology has turned data into an essential component of the production process. Data are automatically generated by machines during the course of operation and during interactions with humans. This paper describes the concept and composition of big data. Most of the big data are unstructured and include text, audio-video files, images, emails, log files, etc. Statisticians are more interested in structured data presented in a pre-defined database model. Big data offer new sources and opportunities that cannot be discounted. However, the use of big data requires proper assessment in terms of quality dimensions such as accuracy, comparability and methodological soundness. Against the backdrop of arguments regarding big data, some users view big data as a replacement of official statistics. Such a conclusion is premature for at least two reasons: first, only a small part of big data can be used for decision-making. Second, theory and practice prove that a small sample based on scientific methods can yield much more reliable and accurate estimates than the results obtained from the processing of large amounts of unstructured data. The paper assesses the possibility of using big data for Sustainable Development Goals (SDG) monitoring, which is a nationally owned process, and NSOs are accountable for the SDG data they report. If the data are derived from a big data source, irrespective of the level of technical sophistication used in data transformation, the reliability of such data might be questioned by the national institutions. The paper concludes that the reliability of data obtained from big data sources hinges on the quality of tools and methods applied to data transformation. Statisticians can play an important role in alerting society, decision-making bodies of the government and businesses about the reliability of information derived from the different sources.

Exploiting Dispersive Power Gain and Delay Spread for Sybil Detection in Industrial WSNs: A Multi-Kernel Approach

Industrial wireless sensor networks (IWSNs) promote innovations in the industry such as structural status mapping, instrument fault diagnosing, and oriented automation system associating. However, due to the shared nature of the wireless propagation environment, the emerging sensor nodes (SNs) with wireless properties are vulnerable to external malicious attacks. The security threats, especially Sybil attacks, impose great difficulties in fulfilling quality requirements of industrial applications. What is more complicated is that the harsh industrial environment brings about new challenges which can degrade the accuracy of detecting Sybil threats. In this paper, we focus on how to detect the malicious packets transmitted from Sybil attackers without adding extra authentication overhead into the transmission frame. We develop a multi-Kernel-based expectation maximization (MKEM) scheme to detect Sybil attacks in IWSNs. Instead of directly investigating the radio resource of SNs, we produce channel-vectors which are extracted from the power gain and delay spread of the channel impulse response obtained from the received packets to represent each SN. Specifically, a kernel-oriented method is designed to discriminate the malicious packets from benign ones without establishing a pre-defined database of channel features of all SNs. Meanwhile, we allocate different kernel weights to the proposed kernels and combine them to improve the discrimination ability of the scheme. Moreover, a kernel parameter optimization method is developed to regulate each kernel weight and parameter to reduce the effects of transmission impairments in IWSNs. To avoid poor detection accuracy when the number of Sybil attackers increases, we use the gap statistical analysis method to verify and EM method to summarize the detection results. The simulation results show that the proposed MKEM scheme can achieve high accuracy on detecting malicious packets transmitted from Sybil attackers from benign ones, and tolerate the effects of transmission impairments in the industrial environment. Moreover, the MKEM scheme can guarantee the detection accuracy even if the number of Sybil attackers increases.

Model-matching for impulse sound localization in urban areas

The presentation addresses the localization of a punctual, impulse sound source with distributed sensors in an urban area of typical size 150mx150m. In the considered approach, the source is localized by matching some observed characteristics of the signals (here, the first times-of-arrival) to those obtained from a pre-defined database of simulations with known source positions. The localization performance is analyzed on the basis of small-scale model and real-scale measurements, with various building heights, source positions, combinations from 15 down to 5 microphones. The time matching localization attains an accuracy of 10 m in the vast majority of the configurations. A confidence level for each localization is satisfactorily tested. The robustness to the buildings height and to the urban map is evaluated. Adaptations of the approach to real-time constraints and to shot and shooter localizations are demonstrated.

Impulse source localization in an urban environment: Time reversal versus time matching

This study investigates two approaches for localizing an impulse sound source with distributed sensors in an urban environment under controlled processing time. In both approaches, the numerical model used for calculating the sound propagation is a finite-difference time-domain (FDTD) model. The simulations are drastically accelerated by restricting to the lower frequencies of the impulse signals and are evaluated against in situ measurements. The first tested localization technique relies on the time reversal of the measurements with the model. In the second technique, the source is localized by matching the observed differences in the first times of arrival of the signals to those obtained from a pre-defined database of simulations with known source positions. The localization performance is physically investigated on the basis of the measurements, considering two source positions and all possible combinations from 5 to 15 microphones. The time matching localization attains an accuracy of 10 m, which is targeted in this study, in the vast majority of the configurations. In comparison, the time reversal localization is affected by the weakness of contributions from sensors masked and distant from the source. Practical requirements are also discussed, such as real-time constraints, hardware and description of the urban environment.

ShrimpGPAT: a gene and protein annotation tool for knowledge sharing and gene discovery in shrimp.

BackgroundAlthough captured and cultivated marine shrimp constitute highly important seafood in terms of both economic value and production quantity, biologists have little knowledge of the shrimp genome and this partly hinders their ability to improve shrimp aquaculture. To help improve this situation, the Shrimp Gene and Protein Annotation Tool (ShrimpGPAT) was conceived as a community-based annotation platform for the acquisition and updating of full-length complementary DNAs (cDNAs), Expressed Sequence Tags (ESTs), transcript contigs and protein sequences of penaeid shrimp and their decapod relatives and for in-silico functional annotation and sequence analysis.DescriptionShrimpGPAT currently holds quality-filtered, molecular sequences of 14 decapod species (~500,000 records for six penaeid shrimp and eight other decapods). The database predominantly comprises transcript sequences derived by both traditional EST Sanger sequencing and more recently by massive-parallel sequencing technologies. The analysis pipeline provides putative functions in terms of sequence homologs, gene ontologies and protein-protein interactions. Data retrieval can be conducted easily either by a keyword text search or by a sequence query via BLAST, and users can save records of interest for later investigation using tools such as multiple sequence alignment and BLAST searches against pre-defined databases. In addition, ShrimpGPAT provides space for community insights by allowing functional annotation with tags and comments on sequences. Community-contributed information will allow for continuous database enrichment, for improvement of functions and for other aspects of sequence analysis.ConclusionsShrimpGPAT is a new, free and easily accessed service for the shrimp research community that provides a comprehensive and up-to-date database of quality-filtered decapod gene and protein sequences together with putative functional prediction and sequence analysis tools. An important feature is its community-based functional annotation capability that allows the research community to contribute knowledge and insights about the properties of molecular sequences for better, shared, functional characterization of shrimp genes. Regularly updated and expanded with data on more decapods, ShrimpGPAT is publicly available at http://shrimpgpat.sc.mahidol.ac.th/.