The early detection of Diabetic Retinopathy (DR) is critical for diabetics to lower the blindness risks. Many studies represent that Deep Convolutional Neural Network (CNN) based approaches are effective enable automatic DR through classifying retinal images patients. Such usually depend on a very large dataset composed with predefined classification labels support their CNN training. However, in some occasions, it not so easy get enough well-labelled act as model training samples. At same...

Modern database management systems (DBMS) expose hundreds of configurable knobs to control system behaviours. Determining the appropriate values for these improve DBMS performance is a long-standing problem in community. As there an increasing number tune and each knob could be continuous or categorical values, manual tuning becomes impractical. Recently, automatic using machine learning methods have shown great potentials. However, existing approaches still incur significant costs only...

Selecting appropriate values for the configurable knobs of Database Management Systems (DBMS) is crucial to improve performance. But because such complexity has surpassed abilities even best human experts, database community turns machine learning (ML)-based automatic tuning systems. However, these systems still incur significant costs or only yield sub-optimal performance, attributable their overly high reliance on black-box optimization and an oversight domain knowledge. This paper...

Selecting appropriate values for the configurable knobs of Database Management Systems (DBMS) is essential to improve performance. But because complexity this task has surpassed abilities even best human experts, database community turns machine learning (ML)- based automatic tuning systems. However, these systems still incur significant costs or only yield suboptimal performance, attributable their overly high reliance on black-box optimization and lack integration with domain knowledge,...

In a research community, the provenance sharing of scientific workflows can enhance distributed cooperation, experiment reproducibility verification and repeatedly doing. Considering that scientists in such community are often loose relation geographically, traditional centralized architectures have shown their disadvantages poor trustworthiness, reliabilities efficiency. Additionally, they also difficult to protect rights interests data providers. All these been largely hindering willings...

A boiler heating surface is composed of hundreds tubes, whose temperatures may be different because their positions, the influences attempering water, and flue gas. Using a criteria based on Davies–Bouldin index, in this article, we propose to partition into local ones, interactions temperature are represented as weighted graph (HSG) at each point time, current features embedded HSG's nodes. Then, prediction model convolutional networks gated recurrent units (WGCN-GRU), proposed. Graph...

Abstract In the traditional medical system, individual data is managed by hospitals rather than patients. It difficult to exchange effectively with fragmented storage, and large amounts of are realize their potential value. With rapid development informatization, centralized storage has been unable meet relevant needs industry. To solve difficulty sharing complexity confirming rights in this paper proposes a model based on blockchain. The provides reliable IPFS file uses Proxy re-encryption...

In foundation pit engineering, the deformation prediction of adjacent pipelines is crucial for construction safety. Existing approaches depend on constitutive models, grey correlation prediction, or traditional feedforward neural networks. Due to complex hydrological and geological conditions, as well nonstationary nonlinear characteristics monitoring data, this problem remains a challenge. By formulating points multivariate time series, deep learning-based model proposed, which utilizes...

The uncertain and diversified nature of scientific research often prohibits scientists from pre-defining a common full-fledged workflow. It would help if end users are allowed to individually construct execute personal workflows on the basis skeleton. In such workflow, can do experiments in "trail-and-error" manner, keep provenance information for future analysis reuse. Driven by these wishes, we proposed an approach workflow developed supporting system called VINCA4Science. A synthesized...

Abstract With the rapid development and wide applications of information technology in medical field, data sharing has become one focus that been received much attention by most researchers recent years. At present, scheme based on blockchain becoming more mature, it features decentralized, secure tamper-resistant to address problem security process sharing, thereby improving quality service citizens, reducing cost cutting down risk. Obviously, is not main factor hinder between institutions....

The outlier detection of high-dimensional data is still challenge. performance existing unsupervised approaches will be affected with the increase outliers in a dataset. stacked autoencoder and GMM are introduced to outliers, an approach termed SAGMM proposed. can reduce reconstruction error observations, determines based on mixture distributions observations obtained model training. Experiments public datasets show that proposed outperforms similar precision, has good balance between...

Modern database management systems (DBMS) expose hundreds of configurable knobs to control system behaviours. Determining the appropriate values for these improve DBMS performance is a long-standing problem in community. As there an increasing number tune and each knob could be continuous or categorical values, manual tuning becomes impractical. Recently, automatic using machine learning methods have shown great potentials. However, existing approaches still incur significant costs only...

Both improving the execution efficiency and reducing cost are essential for scientific workflows in cloud environments. As many workflow tasks become more data-intensive computation-intensive, storing their outputs reuse is a feasible way to achieve such objectives. Because data storage would increase cost, though it might mean less computation time due reuse, important determine proportion of output sets that should be stored cloud. The paper explores caching provenance enhance smart re-run...

In addition to enabling controlled sharing and isolation, the virtualization mechanism can help provide users a single image, realize more abstract uniform view from set of fine-grained, heterogeneous un-ordered instances. The paper explores ways business-oriented service abstraction, Web virtualization, presents practical approach discusses effects thereof in an e-science environment for bioinformatics research.

Outlier detection is essential in many data mining tasks. For high-dimensional data, its outlier often faces two challenges caused by sparse spatial distribution of and big difficulties to get enough class labels. Therefore, it valuable explore a simpler more effective approach unsupervised detection. In this paper, focusing on an based autoencoders Robust PCA proposed. Because PAC has greater advantages feature extraction autoencoder powerful capabilities the reconstruction normal proposed...

Training Back-Propagation Neural Networks (BPNNs) on big datasets faces two challenges, the hight time cost and possibility of getting trapped into local optimum. MapReduce has been introduced to improve efficiency BPNN training in recent years. After each turn split dataset concurrently, lots BPNNs that are only convergent specific will be produced, a global candidate whole needs generated from them. This process is full challenges because it high impact as well accuracy. The paper...

The k-means algorithm is characterized by simple implementation and fast speed, the most widely used clustering algorithm. Aiming at shortcomings of in noise sensitivity high-dimensional sparse data sets, IB (Interpolation-based clustering) proposed. Based on algorithm, genetic for interpolation, which solves problem that easy to merge. experimental results show compared with several improved k-means-based methods, proposed method can achieve better effect deal data.

An infrared image segmentation model for photovoltaic arrays is proposed based on the improved Segformer. The inception-enhanced attention mechanism and multi-scale spatial feature extraction leveraged to address problems, such as holes environmental misclassification. Furthermore, encoder of using Feature Pyramid Network bilinear interpolation operation enhance completeness edge details. Experiments images gathered from a real-world power station shows that achieves improvements 0.48 in...