Fast circular shapes detection in cylindrical ECT sensors by design selection and nonlinear black-box modeling

2017 ◽  
Vol 36 (1) ◽  
pp. 2-17 ◽  
Author(s):  
Yacine Oussar ◽  
Cedric Margo ◽  
Jérôme Lucas ◽  
Stéphane Holé
Keyword(s):  
Cost Effective ◽  
Black Box ◽  
Least Square ◽  
Support Vector ◽  
Content Type ◽  
Circular Shape ◽  
Box Models ◽  
Satisfactory Precision ◽  
Design Selection ◽  
Capacitance Tomography

Purpose Within the framework of image reconstruction in cylindrical electrical capacitance tomography (ECT) sensors, the purpose of this study is to select the structure of a sensor in terms of number and size of the electrodes, to predict the radius and the position of a single circular shape lying in the cross-section defined by the sensor electrodes. Design/methodology/approach Nonlinear black-box models using a set of physically independent capacitances and least-square support vector machines models selected with a sophisticated validation method are implemented. Findings The coordinates of circular shapes are well estimated in fixed and variable permittivity environments even with noisy data. Various numerical experiments are presented and discussed. Sensors formed by three or four electrodes covering 50 per cent of the sensor perimeter provide the best prediction performances. Research limitations/implications The proposed method is limited to the detection of a single circular shape in a cylindrical ECT sensor. Practical implications This method can be advantageously implemented in real-time applications, as it is numerically cost-effective and necessitates a small amount of measurements. Originality/value The contribution is two-fold: a fast computation of a circular shape position and radius with a satisfactory precision compared to the sensor size, and the determination of a cylindrical ECT sensor architecture that allows the most efficient predictions.

scholarly journals Permutation-based Identification of Important Biomarkers for Complex Diseases via Black-box Models

2020 ◽  
Author(s):  
Xinlei Mi ◽  
Baiming Zou ◽  
Fei Zou ◽  
Jianhua Hu
Keyword(s):  
Human Disease ◽  
Molecular Mechanisms ◽  
Black Box ◽  
The Cancer Genome Atlas ◽  
Human Diseases ◽  
Support Vector ◽  
Individual Feature ◽  
Box Models ◽  
Feature Importance ◽  
Black Box Models

AbstractStudy of human disease remains challenging due to convoluted disease etiologies and complex molecular mechanisms at genetic, genomic, and proteomic levels. Many machine learning-based methods, including deep learning and random forest, have been developed and widely used to alleviate some analytic challenges in complex human disease studies. While enjoying the modeling flexibility and robustness, these model frameworks suffer from non-transparency and difficulty in interpreting the role of each individual feature due to their intrinsic black-box natures. However, identifying important biomarkers associated with complex human diseases is a critical pursuit towards assisting researchers to establish novel hypotheses regarding prevention, diagnosis and treatment of complex human diseases. Herein, we propose a Permutation-based Feature Importance Test (PermFIT) for estimating and testing the feature importance, and for assisting interpretation of individual feature in various black-box frameworks, including deep neural networks, random forests, and support vector machines. PermFIT (available at https://github.com/SkadiEye/deepTL) is implemented in a computationally efficient manner, without model refitting for each permuted data. We conduct extensive numerical studies under various scenarios, and show that PermFIT not only yields valid statistical inference, but also helps to improve the prediction accuracy of black-box models with top selected features. With the application to the Cancer Genome Atlas (TCGA) kidney tumor data and the HITChip atlas BMI data, PermFIT clearly demonstrates its practical usage in identifying important biomarkers and boosting performance of black-box predictive models.

Explainable and Interpretable Anomaly Detection Models for Production Data

SPE Journal ◽  
2021 ◽  
pp. 1-15
Author(s):  
Basma Alharbi ◽  
Zhenwen Liang ◽  
Jana M. Aljindan ◽  
Ammar K. Agnia ◽  
Xiangliang Zhang
Keyword(s):  
Anomaly Detection ◽  
Global Analysis ◽  
Black Box ◽  
Prediction Performance ◽  
Support Vector ◽  
Production Data ◽  
K Nearest Neighbor ◽  
Data Set ◽  
Box Models ◽  
Black Box Models

Summary Trusting a machine-learning model is a critical factor that will speed the spread of the fourth industrial revolution. Trust can be achieved by understanding how a model is making decisions. For white-box models, it is easy to “see” the model and examine its prediction. For black-box models, the explanation of the decision process is not straightforward. In this work, we compare the performance of several white- and black-box models on two production data sets in an anomaly detection task. The presence of anomalies in production data can significantly influence business decisions and misrepresent the results of the analysis, if not identified. Therefore, identifying anomalies is a crucial and necessary step to maintain safety and ensure that the wells perform at full capacity. To achieve this, we compare the performance of K-nearest neighbor (KNN), logistic regression (Logit), support vector machines (SVMs), decision tree (DT), random forest (RF), and rule fit classifier (RFC). F1 and complexity are the two main metrics used to compare the prediction performance and interpretability of these models. In one data set, RFC outperformed the remaining models in both F1 and complexity, where F1 = 0.92, and complexity = 0.5. In the second data set, RF outperformed the rest in prediction performance with F1 = 0.84, yet it had the lowest complexity metric (0.04). We further analyzed the best performing models by explaining their predictions using local interpretable model-agnostic explanations, which provide justification for decisions made for each instance. Additionally, we evaluated the global rules learned from white-box models. Local and global analysis enable decision makers to understand how and why models are making certain decisions, which in turn allows trusting the models.

Continuous wide spectrum odor sensing for electronic nose system

Sensor Review ◽  
2018 ◽  
Vol 38 (2) ◽  
pp. 223-230
Author(s):  
Wenli Zhang ◽  
Fengchun Tian ◽  
An Song ◽  
Zhenzhen Zhao ◽  
Youwen Hu ◽  
...  
Keyword(s):  
Support Vector Machine ◽  
Nearest Neighbor ◽  
Wide Spectrum ◽  
Least Square ◽  
Support Vector ◽  
Detection Accuracy ◽  
K Nearest Neighbor ◽  
Sensing Element ◽  
Content Type ◽  
System Errors

Purpose This paper aims to propose an odor sensing system based on wide spectrum for e-nose, based on comprehensive analysis on the merits and drawbacks of current e-nose. Design/methodology/approach The wide spectral light is used as the sensing medium in the e-nose system based on continuous wide spectrum (CWS) odor sensing, and the sensing response of each sensing element is the change of light intensity distribution. Findings Experimental results not only verify the feasibility and effectiveness of the proposed system but also show the effectiveness of least square support vector machine (LSSVM) in eliminating system errors. Practical implications Theoretical model of the system was constructed, and experimental tests were carried out by using NO2 and SO2. System errors in the test data were eliminated using the LSSVM, and the preprocessed data were classified by euclidean distance to centroids (EDC), k-nearest neighbor (KNN), support vector machine (SVM), LSSVM, respectively. Originality/value The system not only has the advantages of current e-nose but also realizes expansion of sensing array by means of light source and the spectrometer with their wide spectrum, high resolution characteristics which improve the detection accuracy and realize real-time detection.

Categorization of fabric design using multi-class least-square support vector machine

2014 ◽  
Vol 26 (1) ◽  
pp. 58-66 ◽  
Author(s):  
A. Ghosh ◽  
T. Guha ◽  
R. Bhar
Keyword(s):  
Pattern Recognition ◽  
Design Methodology ◽  
Inequality Constraints ◽  
Recognition System ◽  
Least Square ◽  
Support Vector ◽  
Content Type ◽  
Reasonable Degree ◽  
Vector Machines ◽  
Fabric Design

Purpose – The purpose of this paper is to give an approach for categorization of diverse textile designs using their textural features as extracted from their gray images by means of multi-class least-square support vector machines (LS-SVM). Design/methodology/approach – In this work, the authors endeavor to devise a pattern recognition system based on LS-SVM which performs a multi-class categorization of three basic woven designs namely plain, twill and sateen after analyzing their features. Findings – The result establishes that LS-SVM is able to classify the fabric design with a reasonable degree of accuracy and it outperforms the standard SVM. Originality/value – The algorithmic simplicity of LS-SVM resulting from replacement of inequality constraints by equality ones and ability of handling noisy data by accommodating an error variable in its algorithm make it eminently suitable for textile pattern recognition. This paper offers a maiden application of LS-SVM in textile pattern recognition.

Spectroscopy quantitative analysis cotton content of blend fabrics

2016 ◽  
Vol 28 (1) ◽  
pp. 65-76 ◽  
Author(s):  
Xudong Sun ◽  
Mingxing Zhou ◽  
Yize Sun
Keyword(s):  
Near Infrared ◽  
Multivariate Calibration ◽  
Partial Least Square ◽  
Least Square ◽  
Support Vector ◽  
Content Type ◽  
Svm Model ◽  
Calibration Methods ◽  
Learning Machine ◽  
Mean Square Errors

Purpose – The purpose of this paper is to develop near infrared (NIR) techniques coupled with multivariate calibration methods to rapid measure cotton content in blend fabrics. Design/methodology/approach – In total, 124 and 41 samples were used to calibrate models and assess the performance of the models, respectively. Multivariate calibration methods of partial least square (PLS), extreme learning machine (ELM) and least square support vector machine (LS-SVM) were employed to develop the models. Through comparing the performance of PLS, ELM and LS-SVM models with new samples, the optimal model of cotton content was obtained with LS-SVM model. The correlation coefficient of prediction (r p ) and root mean square errors of prediction were 0.98 and 4.50 percent, respectively. Findings – The results suggest that NIR technique combining with LS-SVM method has significant potential to quantitatively analyze cotton content in blend fabrics. Originality/value – It may have commercial and regulatory potential to avoid time consuming work, costly and laborious chemical analysis for cotton content in blend fabrics.

Variable snap-off study for 3D additive screen printing

2020 ◽  
Vol 32 (3) ◽  
pp. 430-445
Author(s):  
Yacheng Wang ◽  
Peibo Li ◽  
Yuegang Liu ◽  
Yize Sun ◽  
Liuyuan Su
Keyword(s):  
Theoretical Model ◽  
Regression Model ◽  
Design Methodology ◽  
Screen Printing ◽  
Least Square ◽  
Support Vector ◽  
Content Type ◽  
Printing Quality ◽  
Comparison Results ◽  
The Relationship

Purpose In 3D additive screen printing with constant snap-off, the inhomogeneous screen counterforce will influence the printing force and reduce the printing quality. The purpose of this paper is to study the relationship between scraper position, snap-off and screen counterforce and develop a variable snap-off curve for 3D additive screen printing to improve the printing quality. Design/methodology/approach An experiment was carried out; genetic algorithm (GA) optimization theoretical model, backpropagation neural network regression model and least square support vector machine regression model were established to study the relationship between scraper position, snap-off and screen counterforce. The absolute errors of counterforce of three models with the experiment results were less than 1.5 N, which was tolerated and the three models were considered valid. The comparison results showed that GA optimization theoretical model performed best. Findings The results suggest that GA optimization theoretical model performed best to represent the relationship, and it was used to develop a variable snap-off curve. With the variable snap-off curve in 3D additive screen printing, the inhomogeneous screen counterforce was weakened and the printing quality was improved. Originality/value In printing production, the variable snap-off curve in 3D additive screen printing helps improve the printing quality; this study is of prime importance to the 3D additive screen printing.

Learning solutions to a Cauchy problem for the modified Helmholtz equations using LS-SVM

2020 ◽  
Vol ahead-of-print (ahead-of-print) ◽  
Author(s):  
Ziku Wu ◽  
Xiaoming Han ◽  
GuoFeng Li
Keyword(s):  
Cauchy Problem ◽  
Least Square ◽  
Support Vector ◽  
Helmholtz Equations ◽  
Content Type ◽  
Mesh Free ◽  
Ill Posed ◽  
Approximate Function ◽  
Modified Helmholtz Equations ◽  
First Time

Purpose The purpose of this paper is to develop a mesh-free algorithm based on the least square support vector machines method for numerical simulation of the modified Helmholtz equations. Design/methodology/approach The proposed method deals with a Cauchy problem for the modified Helmholtz equations. The algorithm converts the problem into a quadratic programming. It can be divided into three steps. First, some training points are allocated. Then, an approximate function is constructed. Finally, the shape parameters are estimated. Findings The proposed method's stability is discussed. Numerical experiments are conducted to check the efficiency of the algorithm. The proposed method is found to feasible for the ill-posed problems of the modified Helmholtz equations. Originality/value The originality lies in that the proposed method is applied to solve the modified Helmholtz equations for the first time, and the expected results are obtained.

Spectral dimensionality reduction for quantitative analysis of cotton content of blend fabrics

2019 ◽  
Vol 31 (3) ◽  
pp. 326-338
Author(s):  
Xudong Sun ◽  
Ke Zhu
Keyword(s):  
Dimensionality Reduction ◽  
Multivariate Calibration ◽  
Nir Spectroscopy ◽  
Partial Least Square ◽  
Least Square ◽  
Support Vector ◽  
Wavelet Coefficients ◽  
Content Type ◽  
Calibration Methods ◽  
Spectral Dimensionality

Purpose The purpose of this paper is to initiate investigations to develop near infrared (NIR) spectroscopy coupled with spectral dimensionality reduction and multivariate calibration methods to rapidly measure cotton content in blend fabrics. Design/methodology/approach In total, 124 and 41 samples were used to calibrate models and assess the performance of the models, respectively. The raw spectra are transformed into wavelet coefficients. Multivariate calibration methods of partial least square (PLS), extreme learning machine (ELM) and least square support vector machine (LS-SVM) were employed to develop the models using 100 wavelet coefficients. Through comparing the performance of PLS, ELM and LS-SVM models with new samples, the optimal model of cotton content was obtained with the LS-SVM model. Findings The correlation coefficient of prediction (rp) and root mean square errors of prediction were 0.99 and 4.37 percent, respectively. The results suggest that NIR spectroscopy, combining with the LS-SVM method, has significant potential to quantitatively analyze cotton content in blend fabrics. Originality/value It may have commercial and regulatory potential to avoid time-consuming work, costly and laborious chemical analysis for cotton content in blend fabrics.

Towards an Inductive Logic Programming Approach for Explaining Black-Box Preference Learning Systems

2020 ◽  
Author(s):  
Fabio A. D'Asaro ◽  
Matteo Spezialetti ◽  
Luca Raggioli ◽  
Silvia Rossi
Keyword(s):  
Logic Programming ◽  
Inductive Logic Programming ◽  
Inductive Logic ◽  
Black Box ◽  
User Preferences ◽  
Programming Approach ◽  
Support Vector ◽  
Original Dataset ◽  
Box Models ◽  
Weak Constraints

In this paper we advocate the use of Inductive Logic Programming as a device for explaining black-box models, e.g. Support Vector Machines (SVMs), when they are used to learn user preferences. We present a case study where we use the ILP system ILASP to explain the output of SVM classifiers trained on preference datasets. Explanations are produced in terms of weak constraints, which can be easily understood by humans. We use ILASP both as a global and a local approximator for SVMs, score its fidelity, and discuss how its output can prove useful e.g. for interactive learning tasks and for identifying unwanted biases when the original dataset is not available. Finally, we highlight directions for further work and discuss relevant application areas.

Sign in / Sign up

Export Citation Format

Share Document