COIL Datasets Research Articles

Multi-scale recurrent transformer model for predicting KSTAR PF superconducting coil temperature

Superconducting coils play a critical role in a superconducting-based nuclear fusion device. As the temperature of superconducting magnets increases with a change in current, it is important to predict their temperature to prevent excessive temperature rise of coils and operate them efficiently. We present multi-scale recurrent transformer system, a deep learning model for forecasting the temperature of superconducting coils. Our system recurrently predicts future temperature data of the superconducting coil using the previous data obtained from a multi-scale Korea Superconducting Tokamak Advanced Research poloidal field coil dataset and latent data calculated from previous time step. We apply a multi-scale temperature downsampling approach in our model to effectively learn both the details and the overall structure of the temperature data. We demonstrate the effectiveness of our model through experiments and comparisons with existing models.

Visual Domain Adaptation through Locality Information

Traditional machine learning methods have always performed the learning tasks solely. The models of these conventional methods have to be built from scratch. The idea of domain adaptation can overcome this learning procedure by adapting the knowledge gained from one domain to another. With datasets from several domains, this research proposes a novel strategy for unsupervised visual domain adaptation. Existing transfer learning approaches try to reduce the domain shift using maximum mean discrepancy metric. We propose a combined framework termed Visual Domain Adaptation through Locality Information (DALI) that lowers the geometrical distance between the domains while probabilistically addressing the shortcomings. The proposed approach employs the input data features to develop a technique that assists in inducing the transfer of information from one domain to another. Extensive experimental findings demonstrate that the proposed method DALI excels over traditional domain adaptation algorithms as well as CNN architectures on a variety of cross-domain object, facial, and digit recognition tasks including Office-Caltech, Office-Home, PIE and COIL datasets. Our proposed model has improved the mean accuracy to 69.69%, exceeding the most recent state-of-the-art techniques when performed a comprehensive analysis on the Office-Home data.

Industrial big data-driven mechanical performance prediction for hot-rolling steel using lower upper bound estimation method

Industrial big data technology has become one of the important driving forces to intelligent manufacturing in the steel industry. In this study, the characteristics of data in steel production are analyzed and an industrial big data platform for steeling process is developed to extract the quality-related parameters. A data-driven approach to construct prediction intervals (PIs) of mechanical performances for hot-rolling strips is proposed to represent the uncertainty and reliability of the prediction results. The proposed method employs a new manifold visualization method, SLISEMAP, to reduce the feature dimensions with interpretability, utilizes lower upper bound estimation (LUBE) method to obtain the PIs, in which the broad learning system (BLS) is used as the basic training network model and the artificial bee colony (ABC) algorithm is applied to optimize the weighting parameters of BLS under the LUBE framework. A hot-rolling steel coil dataset consisting of 39 variables and 1335 coil samples is used to validate the proposed method. Two Delta-based approaches, namely back propagation neural network (BPNN) and extreme learning machine (ELM); and three LUBE-based approaches, namely ABC-BPNN, ABC-ELM, and ABC-support vector regression (SVR) are compared with the proposed method. Results show that the proposed ABC-BLS in LUBE is effective and efficient in constructing the PIs with a higher coverage probability and a narrower interval width.

Industrial IoT-Enabled Prediction Interval Estimation of Mechanical Performances for Hot-Rolling Steel

The prediction of mechanical properties holds the key to ensuring product quality and developing new materials in the steel industry. In this study, an industrial Internet of Things platform is developed to obtain the property-related parameters, and a data-driven approach for estimating the prediction intervals (PIs) of these properties is proposed to address the drawbacks of point prediction methods. By combining an optimized extreme learning machine (ELM) and the delta method, the proposed approach specifically applies the regularization mechanism to improve the generalization and stability of the model; it also uses the artificial bee colony algorithm to optimize the initial weights and bias of the input layer in regularized ELM (RELM). An actual steel coil dataset consisting of 27 input dimensions and 2120 samples was used to validate the proposed method, in comparison with the backpropagation neural network, standard ELM, RELM optimized using the genetic algorithm, and lower upper bound estimation method. Experiment results confirm that the proposed method can obtain PIs with a higher coverage probability, narrower interval width, and smaller calculation error.

Object recognition using Hausdorff distance for multimedia applications

The need for reliable and efficient systems for recognition of object from image is increasing day by day. A partial list of applications that may use such system includes searching and reading in hand written documents, recognizing digit on papers and others. In the existing work, Euclidean distance is used for recognizing object, but some of object it doesn’t work well. The major aim of the work is to introduce new object recognition. So the proposed work recognizing object using a shape context and Hausdorff distance is introduced. The process analyses the layout of the image into digits. In the first step, the shape context is computed for two point set and Hungarian algorithm is used to find the correspondence between two point set. The process evaluates the similarity of the two point set using Hausdorff distance. Finally, the error rate is calculated by considering the affine cost and shape context cost. The algorithm tested using the MNIST, COIL data sets and a private collection of hand written digits and encouraging results were obtained. The error rate is reduced to 0.72%.

Generic Object Recognition with Local Receptive Fields Based Extreme Learning Machine

Abstract Generic object recognition is to classify the object to a generic category. Intra-class variabilities cause big troubles for this task. Traditional methods involve plenty of pre-processing steps, like model construction, feature extraction, etc. Moreover, these methods are only effective for some specific dataset. In this paper, we propose to use local receptive fields based extreme learning machine (ELM-LRF) as a general framework for object recognition. It is operated directly on the raw images and thus suitable for all different datasets. Additionally, the architecture is simple and only requires few computations, as most connection weights are randomly generated. Comparing to state-of-the-art results on NORB, ETH-80 and COIL datasets, it is on par with the best one on ETH-80 and sets the new records for NORB and COIL.

Robust GRAPPA reconstruction using sparse multi-kernel learning with least squares support vector regression

Accuracy of interpolation coefficients fitting to the auto-calibrating signal data is crucial for k-space-based parallel reconstruction. Both conventional generalized autocalibrating partially parallel acquisitions (GRAPPA) reconstruction that utilizes linear interpolation function and nonlinear GRAPPA (NLGRAPPA) reconstruction with polynomial kernel function are sensitive to interpolation window and often cannot consistently produce good results for overall acceleration factors. In this study, sparse multi-kernel learning is conducted within the framework of least squares support vector regression to fit interpolation coefficients as well as to reconstruct images robustly under different subsampling patterns and coil datasets. The kernel combination weights and interpolation coefficients are adaptively determined by efficient semi-infinite linear programming techniques. Experimental results on phantom and in vivo data indicate that the proposed method can automatically achieve an optimized compromise between noise suppression and residual artifacts for various sampling schemes. Compared with NLGRAPPA, our method is significantly less sensitive to the interpolation window and kernel parameters.

An evaluation of chemical shift index-based secondary structure determination in proteins: Influence of random coil chemical shifts*

Random coil chemical shifts are commonly used to detect protein secondary structural elements in chemical shift index (CSI) calculations. Though this technique is widely used and seems reliable for folded proteins, the choice of reference random coil chemical shift values can significantly alter the outcome of secondary structure estimation. In order to evaluate these effects, we present a comparison of secondary structure content calculated using CSI, based on five different reference random coil chemical shift value sets, to that derived from three-dimensional structures. Our results show that none of the reference random coil data sets chosen for evaluation fully reproduces the actual secondary structures. Among the reference values generally available to date, most tend to be good estimators only of helices. Based on our evaluation, we recommend the experimental values measured by Schwarzinger et al.(2000), and statistical values obtained by Lukin et al. (1997), as good estimators of both helical and sheet content.

Shape matching and object recognition using shape contexts

We present a novel approach to measuring similarity between shapes and exploit it for object recognition. In our framework, the measurement of similarity is preceded by: (1) solving for correspondences between points on the two shapes; (2) using the correspondences to estimate an aligning transform. In order to solve the correspondence problem, we attach a descriptor, the shape context, to each point. The shape context at a reference point captures the distribution of the remaining points relative to it, thus offering a globally discriminative characterization. Corresponding points on two similar shapes will have similar shape contexts, enabling us to solve for correspondences as an optimal assignment problem. Given the point correspondences, we estimate the transformation that best aligns the two shapes; regularized thin-plate splines provide a flexible class of transformation maps for this purpose. The dissimilarity between the two shapes is computed as a sum of matching errors between corresponding points, together with a term measuring the magnitude of the aligning transform. We treat recognition in a nearest-neighbor classification framework as the problem of finding the stored prototype shape that is maximally similar to that in the image. Results are presented for silhouettes, trademarks, handwritten digits, and the COIL data set.