Boosted Stacking Ensemble Machine Learning Method for Wafer Map Pattern Classification

Recently,machine learning-based technologies have been developed to automate the classification of wafer map defect patterns during semiconductormanufacturing.The existing approaches used in the wafer map pattern classification include directly learning the image through a convolution neural network and applying the ensemble method after extracting image features.This study aims to classify wafer map defects more effectively and derive robust algorithms even for datasets with insufficient defect patterns.First,the number of defects during the actual process may be limited.Therefore,insufficient data are generated using convolutional auto-encoder(CAE),and the expanded data are verified using the evaluation technique of structural similarity index measure(SSIM).After extracting handcrafted features,a boosted stacking ensemble model that integrates the four base-level classifiers with the extreme gradient boosting classifier as a meta-level classifier is designed and built for training the model based on the expanded data for final prediction.Since the proposed algorithm shows better performance than those of existing ensemble classifiers even for insufficient defect patterns,the results of this study will contribute to improving the product quality and yield of the actual semiconductor manufacturing process.

Multi-channel electromyography pattern classification using deep belief networks for enhanced user experience

An enhanced algorithm is proposed to recognize multi-channel electromyography(EMG) patterns using deep belief networks(DBNs). It is difficult to classify the EMG features because an EMG signal has nonlinear and time-varying characteristics.Therefore, in several previous studies, various machine-learning methods have been applied. A DBN is a fast, greedy learning algorithm that can find a fairly good set of weights rapidly, even in deep networks with a large number of parameters and many hidden layers. To evaluate this model, we acquired EMG signals, extracted their features, and then compared the model with the DBN and other conventional classifiers. The accuracy of the DBN is higher than that of the other algorithms. The classification performance of the DBN model designed is approximately 88.60%. It is 7.55%(p=9.82×10-12) higher than linear discriminant analysis(LDA) and 2.89%(p=1.94×10-5) higher than support vector machine(SVM). Further, the DBN is better than shallow learning algorithms or back propagation(BP), and this model is effective for an EMG-based user-interfaced system.

Novel magnetic field computation model in pattern classification

Field computation, an emerging computation technique, has inspired passion of intelligence science research. A novel field computation model based on the magnetic field theory is constructed. The proposed magnetic field computation （MFC） model consists of a field simulator, a non-derivative optimization algo- rithm and an auxiliary data processing unit. The mathematical model is deduced and proved that the MFC model is equivalent to a quadratic discriminant function. Furthermore, the finite element prototype is derived, and the simulator is developed, combining with particle swarm optimizer for the field configuration. Two benchmark classification experiments are studied in the numerical experiment, and one notable advantage is demonstrated that less training samples are required and a better generalization can be achieved.

Encoding candlesticks as images for pattern classification using convolutional neural networks

Candlestick charts display the high,low,opening,and closing prices in a specific period.Candlestick patterns emerge because human actions and reactions are patterned and continuously replicate.These patterns capture information on the candles.According to Thomas Bulkowski’s Encyclopedia of Candlestick Charts,there are 103 candlestick patterns.Traders use these patterns to determine when to enter and exit.Candlestick pattern classification approaches take the hard work out of visually identifying these patterns.To highlight its capabilities,we propose a two-steps approach to recognize candlestick patterns automatically.The first step uses the Gramian Angular Field(GAF)to encode the time series as different types of images.The second step uses the Convolutional Neural Network(CNN)with the GAF images to learn eight critical kinds of candlestick patterns.In this paper,we call the approach GAF-CNN.In the experiments,our approach can identify the eight types of candlestick patterns with 90.7%average accuracy automatically in real-world data,outperforming the LSTM model.

Bridging the Traditional Chinese Medicine Pattern Classification and Biomedical Disease Diagnosis with Systems Biology

Being the unique core of traditional Chinese medicine （TCM）, pattern classification exerts a direct effect on the efficacy and safety of herbal interventions. In this article, the authors integrated the pattern classification and disease diagnosis with many approaches from systems biology, integration of pattern classification with biomedical diagnosis by systems biology is not only a new direction of personalized medicine development, but also provides a new drug development model. In the further study, the pattern classifications of major diseases will be the focus of research.

Pattern Classification of Enterovirus 71-Associated Hand, Foot, and Mouth Disease in Chinese Medicine: A Retrospective Study in 433 Cases

Objective： To determine whether patterns of enterovirus 71（EV71）-associated hand, foot, and mouth disease（HFMD） were classified based on symptoms and signs, and explore whether individual characteristics were correlated with membership in particular pattern. Methods： Symptom-based latent class analysis（LCA） was used to determine whether patterns of EV71-HFMD existed in a sample of 433 cases from a clinical data warehouse system. Logistic regression was then performed to explore whether demographic, and laboratory data were associated with pattern membership. Results： LCA demonstrated a two-subgroup solution with an optimal fit, deduced according to the Bayesian Information Criterion minima. Hot pattern（59.1% of all patients） was characterized by a very high fever and high endorsement rates for classical HFMD symptoms（i.e., rash on the extremities, blisters, and oral mucosa lesions）. Non-hot pattern（40.9% of all patients） was characterized by classical HFMD symptoms. The multiple logistic regression results suggest that white blood cell counts and aspartate transaminase were positively correlated with the hot pattern（adjust odds ratio=1.07, 95% confidence interval： 1.006–1.115; adjust odds ratio=1.051, 95% confidence interval： 1.019–1.084; respectively）. Conclusions： LCA on reported symptoms and signs in a retrospective study allowed different subgroups with meaningful clinical correlates to be defined. These findings provide evidence for targeted prevention and treatment interventions.

Deciphering Potential Correlations between New Biomarkers and Pattern Classification in Chinese Medicine by Bioinformatics:Two Examples of Rheumatoid Arthritis

Biological complexity and the need for personalized medicine means that biomarker development has become increasingly challenging.Thus,new paradigms for research need to be created that bring together a different classifier of individuals.One potential solution is collaboration between biomarker development and Chinese medicine pattern classification.In this article,two examples of rheumatoid arthritis are discussed,including a new biomarker candidate casein kinase 2 interacting protein 1(CKIP-1)and a micro RNA 214.The authors obtained a"snapshot"of pattern classification with disease in biomarker identification.Bioinformatics analyses revealed underlying biological functions of two biomarker candidates,in varying degrees,are correlated with Chinese medicine pattern of rheumatoid arthritis.The authors'initial attempt can provide a new window for studying the win-win potential correlation between the biomarkers and pattern classification in Chinese medicine.

Classification of Short Time Series in Early Parkinson’s Disease With Deep Learning of Fuzzy Recurrence Plots

There are many techniques using sensors and wearable devices for detecting and monitoring patients with Parkinson’s disease(PD).A recent development is the utilization of human interaction with computer keyboards for analyzing and identifying motor signs in the early stages of the disease.Current designs for classification of time series of computer-key hold durations recorded from healthy control and PD subjects require the time series of length to be considerably long.With an attempt to avoid discomfort to participants in performing long physical tasks for data recording,this paper introduces the use of fuzzy recurrence plots of very short time series as input data for the machine training and classification with long short-term memory(LSTM)neural networks.Being an original approach that is able to both significantly increase the feature dimensions and provides the property of deterministic dynamical systems of very short time series for information processing carried out by an LSTM layer architecture,fuzzy recurrence plots provide promising results and outperform the direct input of the time series for the classification of healthy control and early PD subjects.

Ignition Pattern Analysis for Automotive Engine Trouble Diagnosis Using Wavelet Packet Transform and Support Vector Machines

Engine spark ignition is an important source for diagnosis of engine faults.Based on the waveform of the ignition pattern,a mechanic can guess what may be the potential malfunctioning parts of an engine with his/her experience and handbooks.However,this manual diagnostic method is imprecise because many spark ignition patterns are very similar.Therefore,a diagnosis needs many trials to identify the malfunctioning parts.Meanwhile the mechanic needs to disassemble and assemble the engine parts for verification.To tackle this problem,an intelligent diagnosis system was established based on ignition patterns.First,the captured patterns were normalized and compressed.Then wavelet packet transform（WPT） was employed to extract the representative features of the ignition patterns.Finally,a classification system was constructed by using multi-class support vector machines（SVM） and the extracted features.The classification system can intelligently classify the most likely engine fault so as to reduce the number of diagnosis trials.Experimental results show that SVM produces higher diagnosis accuracy than the traditional multilayer feedforward neural network.This is the first trial on the combination of WPT and SVM to analyze ignition patterns and diagnose automotive engines.

Study of a Bionic Pattern Classifier Based on Olfactory Neural System

Simulating biological olfactory neural system, KⅢnetwork, which is a high-dimensional chaotic neural network, is designed in this paper. Different from conventional artificial neural network, the KⅢnetwork works in its chaotic trajectory. It can simulate not only the output EEG waveform observed in electrophysiological experiments, but also the biological intelligence for pattern classification. The simulation analysis and application to the recognition of handwriting numerals are presented here. The classification performance of the KⅢnetwork at different noise levels was also investigated.

Diagnostic performance of endoscopic classifications for neoplastic lesions in patients with ulcerative colitis:A retrospective casecontrol study

BACKGROUND It is unclear whether the Japan Narrow-Band Imaging Expert Team(JNET)classification and pit pattern classification are applicable for diagnosing neoplastic lesions in patients with ulcerative colitis(UC).AIM To clarify the diagnostic performance of these classifications for neoplastic lesions in patients with UC.METHODS This study was conducted as a single-center,retrospective case-control study.Twenty-one lesions in 19 patients with UC-associated neoplasms(UCAN)and 23 lesions in 22 UC patients with sporadic neoplasms(SN),evaluated by magnifying image-enhanced endoscopy,were retrospectively and separately assessed by six endoscopists(three experts,three non-experts),using the JNET and pit pattern classifications.The results were compared with the pathological diagnoses to evaluate the diagnostic performance.Inter-and intra-observer agreements were calculated.RESULTS In this study,JNET type 2 A and pit pattern typeⅢ/Ⅳwere used as indicators of low-grade dysplasia,JNET type 2 B and pit pattern typeⅥlow irregularity were used as indicators of highgrade dysplasia to shallow submucosal invasive carcinoma,JNET type 3 and pit pattern typeⅥhigh irregularity/VN were used as indicators of deep submucosal invasive carcinoma.In the UCAN group,JNET type 2 A and pit pattern typeⅢ/Ⅳhad a low positive predictive value(PPV;50.0%and 40.0%,respectively);however,they had a high negative predictive value(NPV;94.7%and 100%,respectively).Conversely,in the SN group,JNET type 2 A and pit pattern typeⅢ/Ⅳhad a high PPV(100%for both)but a low NPV(63.6%and 77.8%,respectively).In both groups,JNET type 3 and pit pattern typeⅥ-high irregularity/VN showed high specificity.The interobserver agreement of JNET classification and pit pattern classification for UCAN among experts were 0.401 and 0.364,in the same manner for SN,0.666 and 0.597,respectively.The intra-observer agreements of JNET classification and pit pattern classification for UCAN among experts were 0.387,0.454,for SN,0.803 and 0.567,respectively.CONCLUSION The accuracy of endoscopic diagnosis using both classifications was lower for UCAN than for SN.Endoscopic diagnosis of UCAN tended to be underestimated compared with the pathological results.

The pattern characteristics of the tendency variations of earth resistivity and its relation to earthquakes

Through systematically summarizing the observational data of earth resistivity during 26 years from nearly a hundred stations in China, the author found that the pattern of the earth resistivity (ρs) tendency variations,based on monthly average data, could be divided into five types, three types of which were defined as anomalous variation, which have different qualitative and quantitative characteristics and different relations with earthquakes as well.The first type of tendency variation called “funnel” is related to strong earthquakes, the Second type called “scoop” has good corresponding relation with moderate earthquakes, and the third type called “tilt” has no relation with earthquakes. Preliminary discussions about the relations between the three types of ρs tendency variation patterns and earthquakes are made in this paper, according to the experimental results of pressed rocks. It is concluded that the different patterns of ρs tendency variation actually reflect the different stress conditions of underground soil-rock layers: the “funnel” type reflects high stress status, the “scoop” type shows moderate stress condition and the “tilt” type is related to stress relief. All of such knowledges mentioned above are very useful in making accurate medium-term earthquake prediction.

Self‐training maximum classifier discrepancy for EEG emotion recognition

Even with an unprecedented breakthrough of deep learning in electroencephalography(EEG),collecting adequate labelled samples is a critical problem due to laborious and time‐consuming labelling.Recent study proposed to solve the limited label problem via domain adaptation methods.However,they mainly focus on reducing domain discrepancy without considering task‐specific decision boundaries,which may lead to feature distribution overmatching and therefore make it hard to match within a large domain gap completely.A novel self‐training maximum classifier discrepancy method for EEG classification is proposed in this study.The proposed approach detects samples from a new subject beyond the support of the existing source subjects by maximising the discrepancies between two classifiers'outputs.Besides,a self‐training method that uses unlabelled test data to fully use knowledge from the new subject and further reduce the domain gap is proposed.Finally,a 3D Cube that incorporates the spatial and frequency information of the EEG data to create input features of a Convolutional Neural Network(CNN)is constructed.Extensive experiments on SEED and SEED‐IV are conducted.The experimental evaluations exhibit that the proposed method can effectively deal with domain transfer problems and achieve better performance.

A structural developmental neural network with information saturation for continual unsupervised learning

In this paper,we propose a structural developmental neural network to address the plasticity‐stability dilemma,computational inefficiency,and lack of prior knowledge in continual unsupervised learning.This model uses competitive learning rules and dynamic neurons with information saturation to achieve parameter adjustment and adaptive structure development.Dynamic neurons adjust the information saturation after winning the competition and use this parameter to modulate the neuron parameter adjustment and the division timing.By dividing to generate new neurons,the network not only keeps sensitive to novel features but also can subdivide classes learnt repeatedly.The dynamic neurons with information saturation and division mechanism can simulate the long short‐term memory of the human brain,which enables the network to continually learn new samples while maintaining the previous learning results.The parent‐child relationship between neurons arising from neuronal division enables the network to simulate the human cognitive process that gradually refines the perception of objects.By setting the clustering layer parameter,users can choose the desired degree of class subdivision.Experimental results on artificial and real‐world datasets demonstrate that the proposed model is feasible for unsupervised learning tasks in instance increment and class incre-ment scenarios and outperforms prior structural developmental neural networks.

Deep Learning Based Attack Detection for Cyber-Physical System Cybersecurity:A Survey

With the booming of cyber attacks and cyber criminals against cyber-physical systems(CPSs),detecting these attacks remains challenging.It might be the worst of times,but it might be the best of times because of opportunities brought by machine learning(ML),in particular deep learning(DL).In general,DL delivers superior performance to ML because of its layered setting and its effective algorithm for extract useful information from training data.DL models are adopted quickly to cyber attacks against CPS systems.In this survey,a holistic view of recently proposed DL solutions is provided to cyber attack detection in the CPS context.A six-step DL driven methodology is provided to summarize and analyze the surveyed literature for applying DL methods to detect cyber attacks against CPS systems.The methodology includes CPS scenario analysis,cyber attack identification,ML problem formulation,DL model customization,data acquisition for training,and performance evaluation.The reviewed works indicate great potential to detect cyber attacks against CPS through DL modules.Moreover,excellent performance is achieved partly because of several highquality datasets that are readily available for public use.Furthermore,challenges,opportunities,and research trends are pointed out for future research.

Fast Training of Support Vector Machines Using Error-Center-Based Optimization

This paper presents a new algorithm for Support Vector Machine (SVM) training, which trains a machine based on the cluster centers of errors caused by the current machine. Experiments with various training sets show that the computation time of this new algorithm scales almost linear with training set size and thus may be applied to much larger training sets, in comparison to standard quadratic programming (QP) techniques.

Bidirectional Automated Branch and Bound Algorithm for Feature Selection

Feature selection is a process where a minimal feature subset is selected from an original feature set according to a certain measure. In this paper, feature relevancy is defined by an inconsistency rate. A bidirectional automated branch and bound algorithm is presented. It is a new complete search algorithm for feature selection, which performs feature deletion and feature addition in parallel. Its bound is determined by inconsistency rate of the original feature set, hence termed as ‘automated’. Experimental study shows that it is fit for feature selection.

A PATTERN RECOGNITION APPROACH FOR TONE CLASSIFICATION OF ISOLATED SYLLABLE IN PUTONGHUA

In this paper ,a new approach of pattern recognition for tone classification of Putonghua Which is important for speech recognition of Putonghua is discribed . In this method , four parameters of the fundamental frequency trajectory are selected based on a large number of statistical experiments . It is assumed that the four parameters satisfy multidimensional Gaussion distribution and a non-Euclidean distance function for each tone class is derived according to the rule of minimum probability of calssification error . the optimal decision results are obtained in a sense of statistics . It is proved that this method provides very satisfactory results by the experiments for speaker-independent tone classification of Putonghua .

Automated Burned Scar Mapping Using Sentinel-2 Imagery

The Sentinel-2 satellites are providing an unparalleled wealth of high-resolution remotely sensed information with a short revisit cycle, which is ideal for mapping burned areas both accurately and timely. However, the high detail and volume of information provided actually encumbers the automation of the mapping process, at least for the level of automation required to map systematically wildfires on a national level. This paper proposes a fully automated methodology for mapping burn scars using Sentinel-2 data. Information extracted from a pair of Sentinel-2 images, one pre-fire and one post-fire, is jointly used to automatically label a set of training patterns via two empirical rules. An initial pixel-based classification is derived using this training set by means of a Support Vector Machine (SVM) classifier. The latter is subsequently smoothed following a multiple spectral-spatial classification (MSSC) approach, which increases the mapping accuracy and thematic consistency of the final burned area delineation. The proposed methodology was tested on six recent wildfire events in Greece, selected to cover representative cases of the Greek ecosystems and to present challenges in burned area mapping. The lowest classification accuracy achieved was 92%, whereas Matthews correlation coefficient (MCC) was greater or equal to 0.85.

Adaptive segmentation based on multi-classification model for dermoscopy images

Segmentation accuracy of dermoscopy images is important in the computer-aided diagnosis of skin cancer and a wide variety of segmentation methods for dermoscopy images have been developed. Considering that each method has its strengths and weaknesses, a novel adaptive segmentation framework based on multi-classification model is proposed for dermoscopy images. Firstly, five patterns of images are summarized according to the factors influencing segmentation. Then the matching relation is established between each image pattern and its optimal segmentationmethod. Next, the given image is classified into one of the five patterns by the multi-classification model based on BP neural network. Finaily, the optimal segmentation method for this image is selected according to the matching relation, and then the image is effectively segmented. Experiments show that the proposed method delivers better accuracy and more robust segmentation results compared with the other seven state-of-the-art methods.