Search

The process operating performance assessment is critical for the smelting process of electro-fused magnesium furnaces to improve quality of the magnesia product and pursue optimal comprehensive economic benefit. This paper proposes a new method of multi-source heterogeneous information deep feature fusion (MSHIDFF) to achieve higher accuracy operating performance assessment in the electro-fused magnesium smelting process. Firstly, we utilize convolutional neural network, bidirectional long short-term memory network and stacked auto-encoder to extract deep features from raw image, sound and current of different performance grades. Furthermore, those multi-source deep features are fused and the softmax regression with attention mechanism is employed to train a neural network classifier for the fused deep features of different performance grades. The simulation results show that the proposed MSHIDFF method obtains the superior assessment accuracy.

Classical algorithms of relay protection construction do not use all available information base and therefore cannot provide the highest possible sensitivity with guaranteed selectivity. These algorithms, as a rule, concentrate different information, as a result of which it is partially lost. For example, the resistance relay operates with complex resistance, that is, two real parameters, although two complex variables – voltage and current – are used to calculate the complex resistance. This paper shows the solution to the problem of classification of power line operating modes using a neural network algorithm. The simplest neural network, a perceptron, is a universal classifier, since a convergence theorem has been proved for it, showing that if a classification exists, a perceptron of sufficient complexity is able to describe it. The statistical and geometrical interpretations of various algorithms are discussed. The dependence of the quality of the classifier’s work on the distribution of precedents in the training sample, on which the training is based, as well as on the structure and parameters of the neural network, is shown. The recognition ability of the neural network classifier, i.e. the ability to distinguish short circuits within the protected zone from short circuits outside the protected zone at different number of precedents in the training sample, is evaluated. The limits of applicability of such algorithms to the task of classification of object operation modes in electric power industry are shown and recommendations for their practical application are formulated. The results obtained indicate the need to develop methods for training classifiers that are based on a source of informative precedents in the form of a simulation model of the object.

We develop and study an automated method for the detection and classification of three types of technological defects in rolled metal products. The method is based on the ResNet50 neural network classifier, which makes it possible to recognize three classes of defects in the images of flat surfaces with a total accuracy of 95.8% on the basis of the analysis of experimental data with unbalanced numbers of images of different types. To train the classifier, we used about 88,000 images. It is shown that the application of the model developed on the basis of the ResNet50 neural network guarantees its excellent productivity, high quality of recognition, high speed, and high accuracy, which turns the proposed classifier into an effective tool for the solution of problems of engineering diagnostics and nondestructive testing aimed at the classification of defects on the surfaces of rolled metal products.

The authors continued their research on the development of an intelligent automatic ships pilot containing a controller based on fuzzy logic. Its features are determined by the optimizer based on a genetic algorithm. It also contains a modular unit of neural network models of ship navigation paths, as well as a neural network classifier. This paper is devoted to the description of a neural network classifier designed to classify the movement patterns of marine vessels to identify the peculiarities of the ship depending on its type and sailing conditions. The introduction of such classifier to an autopilot allows for more precise consideration of multivariate and difficult to formalize factors affecting the vessel while operating, such as varying weather conditions, irregular waves, hydrodynamic characteristics of the vessel, draft, water under the keel, rate of the vessel sailing, etc. The article outlines the technique concerning the development of a neural network classifier and the results of its computer modelling on the example of a refrigerated transport vessel type. The authors used such methods for obtaining and processing findings as spectral estimation, machine learning methods, in particular, neural network technology and computer or simulation modelling.

BACKGROUD AND OBJECTIVE: The control of clinical manifestation of vestibular system relies on an optimal diagnosis. This study aims to develop and test a new automated diagnostic scheme for vestibular disorder recognition. METHODS: In this study we stratify the Ellipse-fitting technique using the Video Nysta Gmographic (VNG) sequence to obtain the segmented pupil region. Furthermore, the proposed methodology enabled us to select the most optimum VNG features to effectively conduct quantitative evaluation of nystagmus signal. The proposed scheme using a multilayer neural network classifier (MNN) was tested using a dataset involving 98 patients affected by VD and 41 normal subjects. RESULTS: The new MNN scheme uses only five temporal and frequency parameters selected out of initial thirteen parameters. The scheme generated results reached 94% of classification accuracy. CONCLUSIONS: The developed expert system is promising in solving the problem of VNG analysis and achieving accurate results of vestibular disorder recognition or diagnosis comparing to other methods or classifiers.

This paper proposes a novel data-based leakage diagnosis method for big datasets, which identifies the leak occurrence, its size, and its location. Different statistical features are used to express the changes in flow and pressure signals at different leakage scenarios. To improve the performances of the leakage diagnosis approach, factor analysis (FA) is employed for dimension reduction purposes. The optimal features of both pressure and flow signals are then fed as input vectors to a deep neural network (DNN) classifier. The proposed leakage diagnosis method has been applied to the first 20 km of the Golkhari-Binak oil pipeline, located in Iran. The leakage isolation accuracy has been compared with some related works. Simulation results show that the proposed method significantly outperforms others with the average correct classification rate (CCR) of about 98%.

Improved identification of bacterial and viral infections would reduce morbidity from sepsis, reduce antibiotic overuse, and lower healthcare costs. Here, we develop a generalizable host-gene-expression-based classifier for acute bacterial and viral infections. We use training data (N = 1069) from 18 retrospective transcriptomic studies. Using only 29 preselected host mRNAs, we train a neural-network classifier with a bacterial-vs-other area under the receiver-operating characteristic curve (AUROC) 0.92 (95% CI 0.90–0.93) and a viral-vs-other AUROC 0.92 (95% CI 0.90–0.93). We then apply this classifier, inflammatix-bacterial-viral-noninfected-version 1 (IMX-BVN-1), without retraining, to an independent cohort (N = 163). In this cohort, IMX-BVN-1 AUROCs are: bacterial-vs.-other 0.86 (95% CI 0.77–0.93), and viral-vs.-other 0.85 (95% CI 0.76–0.93). In patients enrolled within 36 h of hospital admission (N = 70), IMX-BVN-1 AUROCs are: bacterial-vs.-other 0.92 (95% CI 0.83–0.99), and viral-vs.-other 0.91 (95% CI 0.82–0.98). With further study, IMX-BVN-1 could provide a tool for assessing patients with suspected infection and sepsis at hospital admission., Diagnosing acute infections based on transcriptional host response shows promise, but generalizability is wanting. Here, the authors use a co-normalization framework to train a classifier to diagnose acute infections and apply it to independent data on a targeted diagnostic platform.

This paper proposes the multiple-hypotheses image segmentation and feed-forward neural network classifier for food recognition to improve the performance. Initially, the food or meal image is given...

Stress relaxation plays an important role in the design of underground stopes. The aim of this paper is to assess the stope stability in connection with the stress relaxation using a classification...

The paper shows the efficiency of the multi-class recognition of technical states the welded joint tank by using the neural network classifier that based on Probabilistic Neural Network. Implementation of modern multichannel monitoring systems for determination the technical state of spatial objects requires researching of changes of a stress-strain state construction elements which function under the operation pressure and possible impact its structural integrity. Such researching are needed for prevention of cracks or damage determining and future prediction of object’s technical state. In the paper the tasks of a multi-class recognition of a state of multi-site damage the welded joint tank are determined. In the research was created the model of the object with probable location places cracks. One of them is a vertical crack and two are horizontal. Directions of their propagation are given. Cracks have the same value. The first researching included when cracks turn up one by one. The next one related with multi-focal defects when cracks were arised and they evolve in parallel and independently. The data of strain in the structural of the welded joint tank where sensors were attached is given. The research of the possibility of the error-free recognition was conducted by the developed classifier based on the stress-strain state of the geometric model of the tank structural elements with multi-site damage, where sensors are located. The development of the classifier was done by using Probabilistic Neural Network, which provides the best results of a multi-class recognition for determination technical state of spatial object with multidimensional vectors of diagnostic features. As a result, the probability of recognition from the network influence parameter, which shows the effectiveness of the neural network classifier for localization of single damage and localization of multiple cracks, was established

Due to advancement of new edge medical technologies, many methods have been applied to solve medical issues including machine learning approach. Cardiac Arrhythmia is one of the common diseases which can be solved using various machine learning approaches. There are many approaches which have already been introduced to classify arrhythmia and abnormality detection. This paper has a solution, introduces supervised and unsupervised models in which supervised models generate a good classification result. However, in this paper, we have also introduced a deep neural network classifier and used for prediction of arrhythmia if present based on some predefined value. In this paper, we have also connected it to the user interface to which the native users can check the level of arrhythmia.

Background: Medical image segmentation is a well-studied subject within the field of image processing. The goal of this research is to create an AI retinal screening grading system that is both accurate and fast. We introduce a new segmentation network which achieves state-of-the-art results on semantic segmentation of color fundus photographs. By applying the net-work to identify anatomical markers of diabetic retinopathy (DR) and diabetic macular edema (DME), we collect sufficient information to classify patients by grades R0 and R1 or above, M0 and M1. Methods: The AI grading system was trained on screening data to evaluate the presence of DR and DME. The core algorithm of the system is a deep learning network that segments relevant anatomical features in a retinal image. Patients were graded according to the standard NHS Diabetic Eye Screening Program feature-based grading protocol. Results: The algorithm performance was evaluated with a series of 6,981 patient retinal images from routine diabetic eye screenings. It correctly predicted 98.9% of retinopathy events and 95.5% of maculopathy events. Non-disease events prediction rate was 68.6% for retinopathy and 81.2% for maculopathy. Conclusion: This novel deep learning model was trained and tested on patient data from annual diabetic retinopathy screenings can classify with high accuracy the DR and DME status of a person with diabetes. The system can be easily reconfigured according to any grading protocol, without running a long AI training procedure. The incorporation of the AI grading system can increase the graders’ productivity and improve the final outcome accuracy of the screening process.

Computer-assisted colon cancer detection on the histopathological images has become a tedious task due to its shape characteristics and other biological properties. The images acquired through the histopathological microscope may vary in magnifications for better visibility. This may change the morphological properties and hence an automated magnification independent colon cancer detection system is essential. The manual diagnosis of colon biopsy images is subjective, sluggish, laborious leading to nonconformity between histopathologists due to visual evaluation at various microscopic magnifications. Automatic detection of colon across image magnifications is challenging due to many aspects like tailored segmentation and varying features. This demands techniques that take advantage of the textural, color, and geometric properties of colon tissue. This work exhibits a segmentation approach based on the morphological features derived from the segmented region. Gabor Wavelet, Harris Corner, and DWT-LBP coefficients are extracted as it should not be dependent on the spatial domain with respect to the magnification. These features are fed to the Genetically Optimized Neural Network classifier to classify them as normal and malignant ones. Here, the genetic algorithm is used to learn the best hyper-parameters for a neural network.

This paper focuses of my enrollment in ICICS 2021 Competition Mowjaz Multi-Topic Labelling Task using Bidirectional Gated Recurrent Unit (Bi-GRU). The model is basically used to classify articles based on their topics that are present within its content. Mowjaz’s topic are classified into ten categories and an article can be classified as under as many topics as it covers. In the evaluation, we regard the Mowjaz Multi-Topic labelling task as multi-classification task and use Unigram models to extract features to train a neural network classifier. In the result, the accuracy of my method reached 0.8232, ranking 8th .

Summary Neural networks have shown strong potential in research and in healthcare. Mainly due to the need for large datasets, these applications have focused on common medical conditions, where more data are typically available. Leveraging publicly available data, we trained a neural network classifier on images of rare genetic conditions with skin findings. We used approximately 100 images per condition to classify 6 different genetic conditions. We analyzed both preprocessed images that were cropped to show only the skin lesions as well as more complex images showing features such as the entire body segment, the person, and/or the background. The classifier construction process included attribution methods to visualize which pixels were most important for computer-based classification. Our classifier was significantly more accurate than pediatricians or medical geneticists for both types of images and suggests steps for further research involving clinical scenarios and other applications., We built a neural network classifier for images of skin findings in selected genetic conditions. We compared the accuracy (using both cropped and larger images) of our classifier to that of pediatricians and medical geneticists. We show how attribution methods can help refine a classifier.

In this study, a novel method, fuzzy nearest neighbor partitioning (FNNP), is proposed for improving the classification capability of the neural network. For the original nearest neighbor partitioning method, the “short-sightedness defect” problem restrains the nearest neighbor partitioning method to learn valuable experience from the distribution information outside the samples’ nearest neighbors. Moreover, the noise sensitive problem that exists in the process of classifying the unknown samples further influences the performance of the nearest neighbor partitioning method. To overcome the short-sightedness defect problem, fuzzy logic theory is integrated into the FNNP, and promotes the ability that the FNNP learns experience from extensive distribution information. In addition, an improved classification strategy, which adopts the concept of the fuzzy nearest neighbor, is employed to increase the immunity of FNNP to noise when the unknown samples are classified. The proposed FNNP is compared with other classification methods for some famous datasets. The results of the experiments indicate that the proposed method achieves promising classification performance for various indicators.

Automated classification and morphological analysis of white blood cells has been addressed since last four decades, but there is no optimal method which can be used as decision support system in laboratories due to biologically complex nature of the cells. Automated blood cell analysis facilitates quick and objective results and can also handle massive amount of data without compromising with efficiency. In the present study, we demonstrate classification of white blood cells into six types namely lymphocytes, monocytes, neutrophils, eosinophils, basophils and abnormal cells. We provide the comparison of traditional image processing approach and deep learning methods for classification of white blood cells. We evaluated neural network classifier results for hand-crafted features and obtained the average accuracy of 99.8%. We also used full training and transfer learning approaches of convolutional neural network for the classification. An accuracy around 99% was obtained for full training CNN.

Character level models are drawing attention recently. A number of these models have been proposed and shown successful in Natural Language Processing tasks. While most of the models are experimented mainly on English, or other alphabetic languages, a number of problems arise when they applied these models to non-alphabetic language such as Chinese. In this study, we investigated the problems encountered when transferring these models to the Chinese and put forward some solutions. We propose a double embedding neural network model that is also character level and consists of both CNN and RNN with two separate embeddings. The model is applied to a fundamental Natural Language Processing task, text classification. Experiment results conducted on the Chinese corpus demonstrated that our character level neural network model performs just as well as or better than those word level classification models. Our model is able to reach 95.9% accuracy on a Chinese Fudan news dataset, which outperforms the state-of-the-art models.

Traditional training algorithms in artificial neural networks (ANNs) show some inherent weaknesses, such as the possibility of falling into local optimum, slow learning speed, and the inability to determine the optimal neuronal structure. To remedy the deficiencies of traditional neural networks, this paper proposes a novel neural network classifier (NNC) using the beetle antennae search (BAS) algorithm, termed BASNNC. The BAS algorithm is explored to optimize the weights of the NNC. The network of the proposed BASNNC adopts three-layer structure, including an input layer, a hidden layer, and an output layer. Quite differing from the traditional training algorithm using a principle of gradient descent, the weights between the hidden and output layers are optimized by the BAS algorithm, which effectively improves the computational speed of the classifier. The numerical studies, applications to pattern classification and comparisons with an error back-propagation neural network model, show that the proposed BASNNC has faster computational speed and higher classification accuracy.

© 2013 IEEE. In this paper, an Advanced Quantum-based Neural Network Classifier (AQNN) is proposed. The proposed AQNN is used to form an objectionable Web content filtering system (OWF). The aim is to design a neural network with a few numbers of hidden layer neurons with the optimal connection weights and the threshold of neurons. The proposed algorithm uses the concept of quantum computing and genetic concept to evolve connection weights and the threshold of neurons. Quantum computing uses qubit as a probabilistic representation which is the smallest unit of information in the quantum computing concept. In this algorithm, a threshold boundary parameter is also introduced to find the optimal value of the threshold of neurons. The proposed algorithm forms neural network architecture which is used to form an objectionable Web content filtering system which detects objectionable Web request by the user. To judge the performance of the proposed AQNN, a total of 2000 (1000 objectionable + 1000 non-objectionable) Website's contents have been used. The results of AQNN are also compared with QNN-F and well-known classifiers as backpropagation, support vector machine (SVM), multilayer perceptron, decision tree algorithm, and artificial neural network. The results show that the AQNN as classifier performs better than existing classifiers. The performance of the proposed objectionable Web content filtering system (OWF) is also compared with well-known objectionable Web filtering software and existing models. It is found that the proposed OWF performs better than existing solutions in terms of filtering objectionable content.

One of the known problems in security systems is to identify persons based on certain signatures. Biometrics have been adopted in security systems to identify persons based on some physiological or behavioral characteristics that they own. Photoplethysmography (PPG) is a physiological signal that is used to describe the volumetric change of blood flow in peripherals with heartbeats. The PPG signals gained some interest of researchers in the last few years, because they are used non-invasively, and they are easily captured by the emerging IoT sensors from fingertips. This paper presents a PPG-based approach to identify persons using a neural network classifier. Firstly, PPG signals are captured from a number of persons using IoT sensors. Then, unique features are extracted from captured PPG signals by estimating the Mel-Frequency Cepstral Coefficients (MFCCs). These features are fed into an Artificial Neural Network (ANN) to be trained first and used for identification of persons. A dataset of PPG signals for 35 healthy persons was collected to test the performance of the proposed approach. Experimental results demonstrate 100% and 98.07% accuracy levels using the hold-out method and the 10-fold cross-validation method, respectively.

Based on machine vision, this paper first gives the overall design of the industrial robot sorting system and the composition of the experimental platform of each subsystem; then constructs the neural network classifier to recognize the target mechanical workpiece, and uses the mean shift iterative algorithm to track the moving target; finally, it expounds the sorting strategy of Huashu industrial robot and completes the sorting experiment.

This paper presents an enhanced pattern recognition neural network classifier for ground target classification in FSR. The hybrid enhanced model is based on a different combination of pre-processing techniques. Wavelet and feature extraction techniques were applied on time-domain raw data of three vehicles before it is processed by the neural network classifier. Several scenarios were undertaken to study and analyze the effect of adding the proposed techniques to the NN classifier’s performance. The result reveals the potential and effectiveness of pre-processing to increase the performance by 79.2% to previous work, which gives more than 97% true classification.

ObjectivesTo evaluate the effectiveness of radiomic features on classifying histological subtypes of central lung cancer in contrast-enhanced CT (CECT) images.Materials and MethodsA total of 200 patients with radiologically defined central lung cancer were recruited. All patients underwent dual-phase chest CECT, and the histological subtypes (adenocarcinoma (ADC), squamous cell carcinoma (SCC), small cell lung cancer (SCLC)) were confirmed by histopathological samples. 107 features were used in five machine learning classifiers to perform the predictive analysis among three subtypes. Models were trained and validated in two conditions: using radiomic features alone, and combining clinical features with radiomic features. The performance of the classification models was evaluated by the area under the receiver operating characteristic curve (AUC).ResultsThe highest AUCs in classifying ADC vs. SCC, ADC vs. SCLC, and SCC vs. SCLC were 0.879, 0.836, 0.783, respectively by using only radiomic features in a feedforward neural network.ConclusionOur study indicates that radiomic features based on the CECT images might be a promising tool for noninvasive prediction of histological subtypes in central lung cancer and the neural network classifier might be well-suited to this task.

Covid19 is a horrible disease, which upset our life everywhere. The main complexity of this disease lies in its rapid evolution and through people’s contact and gaps in our understanding. Moreover, it represents critical cases when the immune system has not presented any symptoms. Hence, the design of an effective classifier is necessary. This paper aims to hybrid the multi-criteria Analytic Hierarchy Pprocess (AHP) tool and the process of the convolutional neural network (CNN), for making the classification of a category of patients. Our novel method is divided into two main phases: the first one focuses on the generation of the priorities of the essential criteria using the AHP model, while the second phase aims to classify the patients using the neural network classifier. In the present study, we considered three important criteria: fever, patient, localiztion, and the age of the patient. From the obtained results, the proposed model has proved its efficiency even if we consider different cases.

A method for disturbance positioning in noisy environment is introduced. By the partitioning algorithm and a well-trained B-P neural network classifier, an average locating accuracy of 95% was demonstrated in the field test.

Recently, analysing reviews presented by clients to products that are provided by e-commerce companies, such as Amazon, to produce efficient recommendations has been receiving a lot of attention. However, ensuring and generating effective recommendations on time is a challenge. This research paper proposes an artificial intelligence-based system. The proposed system uses the Incremental Learning - based Method (ILbM) to learn a neural network classifier. The ILbM uses the bagging technique in the process of training the classifier. To ensure a high degree of performance, the ILbM is implemented on the Hadoop since it allows the execution in parallel. Compared to a similar system, the proposed system shows better results in terms of accuracy (97.5%), precision (95.7%), recall (91.5%), and time of response (36 seconds).

In this study we present a system for automated processing of signals from the polysomnogram (PSG) for the detection of apnoea and non-apnoea arousals. The PSG signals were divided into 15 second epochs and 59 time- and frequency-domain features were derived for each epoch. Features from adjacent 4 epochs were combined and processed with a bank of ten feed-forward neural networks each with a single hidden layer of 20 units. The system outputs a 200 Hz annotation signal containing probability estimates that each sample was associated with an apnoea or non-apnoea arousal, or no-arousal. Data from the Physionet Computing in Cardiology Challenge 2018 was used to develop and test the system. Performance of the system was assessed using three class and two class metrics. With the system classifying three classes, the volume under the receiver operator characteristic (ROC) surface was 0.75 with an optimal specificity of 0.72, a sensitivity of 0.76 for the apnoea arousals, and a sensitivity of 0.69 for the non-apnoea arousals. When the two arousal classes were combined into one arousal class, the area under the precision recall curve was 0.74, the area under the ROC curve was 0.91, with an optimal specificity and sensitivity of 0.85.

Error-Related Potential (Errp) is the bioelectric potential elicited in human brain as a result of the cognitive state of awareness when an error is perceived. Identifying ErrP from a single trial electroencephalogram (EEG) data can be used in taking corrective actions to fix the error or as a learning strategy in Brain Computer Interface (BCI) systems. The ErrP dataset recorded using EEG will contain both erroneous and correct actions. A classifier such as the Artificial Neural Network (ANN) can be trained to identify the erroneous versus correct action from a single trial EEG data. However, the classifier will have large number of parameters to be learned, and typically, the ErrP dataset is unbalanced with smaller number of erroneous trials. Therefore, the trained classifier may not generalize the data well. To classify the ErrP with better accuracy, an ANN architecture is proposed in this work. Learning the parameters of the ANN is carried out in two stages (Stage-1 and Stage-2) in the proposed method. The first stage of learning will have relatively large feature samples collected from several subjects. The first stage learning is aimed to capture the global characteristics of the ErrP. In the second stage, the pre-trained ANN classifier from the first stage is tuned for each subject. The ErrP dataset has two sessions dataset recorded from six subjects and the Stage-1 and Stage-2 training models are cross-validated. The overall classification accuracy achieved after cross-validation is 74.78 ± 3.43% and 86.03 ± 1.02% for erroneous and correct trials respectively. The improvement in the classification accuracy achieved is 12.67% and 15.51% for erroneous and correct trials respectively compared with the existing statistical classifier method. The method proposed shows an efficient way to train ANN classifier to achieve higher classification accuracy for unbalanced and smaller dataset such as ErrP.

A new approach of using the paradigm of “interval teacher” in the classical scheme of training neural networks with a teacher is used to solve the problem of medical diagnosis of the disease coronary heart disease. An NN architecture has been created that implements a mechanism for selecting a specific “teacher” from a deterministic interval set. A comparative analysis of the proposed and classical training schemes of the neural network is shown, confirming the effectiveness of the proposed training concept. The created medical classifier based on a neural network with an interval teacher provides high classification accuracy.

A convolutional neural network (CNN) based classifier, to predict the retweet level of COVID-19 tweets, is proposed in this paper. The proposed CNN is able to predict whether a given COVID-19 tweet would be more retweeted, or less retweeted. The network is trained and validated with 100,000 and 5,000 English tweet samples, respectively, which were all posted within the last week of March 2020, and 81% accuracy has been achieved. The network is also evaluated by English tweet samples posted at the end of April. The result shows that the accuracy is about 80%. Therefore, the proposed approach is robust and capable to process tweets of chosen contents/topics.

The information about intended hand gestures can be extracted by processing surface electromyography signals using non-invasive commercial off the shelf surface electromyography data acquisition devices. Surface electromyography signals have a great potential for use in multi-functional prosthetic controllers. The objective of this study is the implementation of a classifier that can be used to classify gestures from Myo-electric data obtained from the Myo-armband. This study describes in detail a method for data acquisition, feature extraction, and offline gesture classification using Artificial Neural Network. The performance is then compared with a Support Vector Machine Classifier. The proposed approach results in an accuracy greater than 94% for validation data set for classification of five distinct hand gestures. It could be concluded that this technique could be used in the human-machine interfaces with five distinct control signals including rest. A significant observation in this study was that a single artificial neural network taking inputs from all sensors simultaneously gives inferences with better accuracy compared to a system with a separate neural network for each sensor with a majority voting to decide the classification of the gesture.

Research in Machine Learning (ML) for classification tasks has been primarily guided by metrics that derive from a confusion matrix (e.g. accuracy, precision and recall). Several works have highlighted that this has lead to training practices that produce over-confident models and void the assumption that the model learns a probability distribution over the classification targets; this is referred to as miscalibration. Consequently, modern ML architectures struggle to perform in applications where a probabilistic forecaster is needed. Research efforts on calibration techniques have explored the possibility of recovering probability distributions from traditional architectures. This tutorial covers the key concepts required to understand the motivations behind calibration and aims at providing participants with the tools that they require assess the calibration of ML models and calibrate them when required.

Feature extraction is one of the key technologies in pattern recognition, which has great influence on the accuracy of text recognition. In this paper, The Tai Le are extracted with coarse grid features, directional line element feature and penetration number feature, evaluated using BP neural network classifier. The results show that the direction line element feature on the Tai Le print recognition data set YDH2019.1 has higher recognition accuracy than the other two features, is more suitable for the Tai Le recognition.

This paper presents a neural network classifier approach to detecting precise within-document (WD) and cross-document (CD) event coreference clusters effectively using only event mention based features. Our approach does not rely on any event argument features such as semantic roles or spatio-temporal arguments and uses no sophisticated clustering approach. Experimental results on the ECB+dataset show that our simple approach out performsstate-of-the-art methods for both within-document and cross-document event coreference resolution while producing clusters of high precision, which is useful for several downstream tasks.

In the industries of wood furniture and wood floor manufacturing, the color and texture of wood panel are important surface feature parameters. At present, the comprehensive measurement visual feature above-mentioned of wood panel mainly relies on manual visual inspection, which is greatly influenced by human factors, slow in measurement speed and low in accuracy. In this paper, color histogram and gray level co-occurrence matrix are used to digitize the visual features of wood panel, and a parameter system suitable for describing the visual features of wood panel is established. Using by the visual feature parameters system Vnew of wood panel, the recognition rate of BP neural network classifier is up to 95.25%, and the results show that the proposed method is feasible.

Ventricular tachycardia is a life threatening medical emergency. Discerning dangerous ventricular rhythms with safe Sinus tachycardia based on heart rate is very tough as they are having similar heart rate. Most of the existing research used time information for classification which may lead false alarm. Hence a CMOS circuit is proposed to classify ventricular-tachycardia based on morphological changes in QRS complex. The design includes sample and hold circuit for sampling QRS complex, mapping circuit for map the given input signal to unit length, hamming neural network and winner take all circuits for classification of ventricular tachycardia. This design is implemented using 180nm CMOS technology with the operating voltage and power consumption of 19.81µW.

This paper is concerned with the combination of microscopic image processing tools and artificial intelligence in order to detect and recognize the Nosema disease, which affects honey bees. In fact, the use of imaging processing tools in medical and biological sciences was significantly increased, especially in the detection and diagnosis of diseases. Our work focuses on a segmentation method which analyses the objects of an image and decides if it is Nosema or not. A set of the significant values of geometric and texture features has been calculated and fused as a definition for a Nosema cell. A MultiLayer neural network was applied as classifier to detect and recognize this disease on Nosema images. Finally, our automatic segmentation approach shows a 91% of accuracy, in the process of identifying the Nosema disease.

Researchers are continuously making breakthroughs on the impact of deep learning in the medical industry. This approach of Deep Learning (DL) in neuroimaging creates new insights in modification of brain structures during various disorders, helping capture complex relationships that may not have been visible otherwise. The main aim of proposed work is to effectively detect the presence of schizophrenia, a mental disorder that has drastic implications and is hard to spot, using Magnetic Resonance Image Features from fMRI Database. Dataset is then fed to a Neural Network classifier, which learns to predict and give indications for preventing the onset of Schizophrenia using regression models.

Presently, the usage of medicinal images has drastically increased and it provides extensive details related to the patient’s health status. It shows the applicability of diagnosing the disease and stored in a memory for examination purposes. For the retrieval of medical images in a real world environment, significant need is present in the designing of an effective medical image indexing and retrieval technique. This paper offers an efficient medical image retrieval (MIR) model through feature extraction based classification model. Here, Directional local ternary quantized extrema patterns (DLTerQEPs) and autocorrelogram (AC) based feature extraction process takes place to extract texture and color features. Next, neural network (NN) based classification process takes place. The investigation of the simulation results takes place to showcase the betterment of the presented mode. During experimentation process, it is noticed that the presented model is superior to compared methods.

Classification is a machine learning technique that is used to find the membership of an object to a given set of classes or groups. Neural network (NN) classifier always suffers some issues while examining outliers and data points from nearby classes. In this paper, we present a new hyper-credal neural network classifier based on belief theory. This method is based on credal classification technique introduced in Dempster-Shafer Theory (DST). It allows a data point to belong not only to a specific class but also to a meta-class or an ignorant class based on its mass. The sample which lies in an overlapping region is accurately classified in this method to a meta-class which corresponds to the union of the overlapping classes. Therefore this approach reduces the classification error at the price of precision. But this decrease in precision is acceptable in applications such as medical, defence related applications where a wrong decision would cost more than avoiding some correct decisions. The results and analysis of different databases are given to illustrate the potential of this approach. This idea of hyper-credal classification is extended to convolutional neural network (CNN) classifiers also.

The application of machine learning programs is becoming increasingly more widespread. Neural networks, which are among the most popular machine learning programs, play important roles in people's daily lives, such as by controlling cars in autonomous driving systems. However, neural networks still lack effective testing methods. To address this problem, this paper proposes a testing method for neural network classifiers based on metamorphic relations. Firstly, it designs metamorphic relations to transform the original data set into derivative data sets. Then, it uses the data before and after the transformation to train and test the neural network classifier, respectively. Finally, it checks whether the output conforms to the metamorphic relations. The neural network classifier is defective if conflicts are detected. Experiments are conducted on a neural network classifier from Stanford's cs231n course to verify the effectiveness of the method. The results show that the defect detection capability of the proposed method is accurate, and 87.5% of the mutants are successfully detected.