Search

The explosive growth of geospatial data is increasing requirements for automatic and efficient data learning abilities. Many deep learning methods have been widely applied for geospatial data understanding tasks, such as road networks and geospatial object detection. However, the demands for more accurate learning of high-level features require the use of deeper neural networks. To further improve the learning efficiency of deep neural networks, in this paper, we propose an improved convolutional neural network named EDense. First, we use its dense connectivity to integrate a CNN with an extreme learning machine. Then, we expand the kernels in the convolutional layers to increase the width of the network model. Furthermore, we propose one-feature EDense (OF-EDense), which is a simplified version of EDense, to fit conditions in which the number of parameters is strictly limited. Finally, the experimental results fully demonstrate the strong learning ability and high learning efficiency of EDense. [ABSTRACT FROM AUTHOR]

The application of Deep Neural Networks (DNNs) to a broad variety of tasks demands methods for coping with the complex and opaque nature of these architectures. When a gold standard is available, performance assessment treats the DNN as a black box and computes standard metrics based on the comparison of the predictions with the ground truth. A deeper understanding of performances requires going beyond such evaluation metrics to diagnose the model behavior and the prediction errors. This goal can be pursued in two complementary ways. On one side, model interpretation techniques "open the box" and assess the relationship between the input, the inner layers and the output, so as to identify the architecture modules most likely to cause the performance loss. On the other hand, black-box error diagnosis techniques study the correlation between the model response and some properties of the input not used for training, so as to identify the features of the inputs that make the model fail. Both approaches give hints on how to improve the architecture and/or the training process. This paper focuses on the application of DNNs to computer vision (CV) tasks and presents a survey of the tools that support the black-box performance diagnosis paradigm. It illustrates the features and gaps of the current proposals, discusses the relevant research directions and provides a brief overview of the diagnosis tools in sectors other than CV. [ABSTRACT FROM AUTHOR]

In machine learning models, one of the most popular models is artificial neural networks. The activation function is one of the important parameters of neural networks. In this paper, the sigmoid function is used as an activation function with a fractional derivative approach to minimize the convergence error in backpropagation and to maximize the generalization performance of neural networks. The proportional Caputo definition is considered a fractional derivative. We evaluated three neural network models on the usage of the proportional Caputo derivative. The results show that the proportional Caputo derivative approach has higher classification accuracy than traditional derivative models in backpropagation for neural networks with and without L2 regularization. [ABSTRACT FROM AUTHOR]

Deep learning techniques, in particular generative models, have taken on great importance in medical image analysis. This paper surveys fundamental deep learning concepts related to medical image generation. It provides concise overviews of studies which use some of the latest state-of-the-art models from last years applied to medical images of different injured body areas or organs that have a disease associated with (e.g., brain tumor and COVID-19 lungs pneumonia). The motivation for this study is to offer a comprehensive overview of artificial neural networks (NNs) and deep generative models in medical imaging, so more groups and authors that are not familiar with deep learning take into consideration its use in medicine works. We review the use of generative models, such as generative adversarial networks and variational autoencoders, as techniques to achieve semantic segmentation, data augmentation, and better classification algorithms, among other purposes. In addition, a collection of widely used public medical datasets containing magnetic resonance (MR) images, computed tomography (CT) scans, and common pictures is presented. Finally, we feature a summary of the current state of generative models in medical image including key features, current challenges, and future research paths. [ABSTRACT FROM AUTHOR]

Demand forecasting is a scientific and methodical assessment of future demand for a critical product.The effective Demand Forecast Model (DFM) enables pharmaceutical companies to be successful in the global market. The purpose of this research paper is to validate various shallow and deep neural network methods for demand forecasting, with the aim of recommending sales and marketing strategies based on the trend/seasonal effects of eight different groups of pharmaceutical products with different characteristics. The root mean squared error (RMSE) is used as the predictive accuracy of DFMs. This study also found that the mean RMSE value of the shallow neural network-based DFMs was 6.27 for all drug categories, which was lower than deep neural network models. According to the findings, DFMs based on shallow neural networks can effectively estimate future demand for pharmaceutical products. [ABSTRACT FROM AUTHOR]

In this paper, we propose a method to approximate the solutions to nonlinear ordinary differential equations (ODE) using a deep learning feedforward artificial neural networks (ANNs). The efficiency of the proposed—unsupervised type machine learning—method is shown by solving two boundary value problems (BVPs) from quantum mechanics and nanofluid mechanics. The proposed mean-squared loss function is the sum of two terms: the first term satisfies the differential equation, while the second term satisfies the initial or boundary conditions. The total loss function is minimized by using general type of quasi-Newton optimization methods to get a desired network output. The approximation capability of the proposed method is verified for two sets of boundary value problems: first, a second-order nonlinear ODE and, second, a system of coupled nonlinear third-order ODEs. Point-wise comparison of our approximation shows a strong agreement with the available exact solutions and/or Runge–Kutta-based numerical solutions. We remark that the proposed algorithm minimizes the overall learnable network hyperparameters in a given initial or boundary value problems. More importantly, for the coupled system of third-order nonlinear ordinary differential equations, the proposed method does not need any adjustment with the initial/boundary conditions. Also, the current method does not require any special type of computational mesh. A straightforward minimization of total loss function yields a highly accurate results even with less number of epochs. Therefore, the proposed framework offers an attractive setting for the fluid mechanics community who are interested in studying heat and mass transfer problems. [ABSTRACT FROM AUTHOR]

As a compact and effective learning model, the random vector functional link neural network (RVFL) has been confirmed with universal approximation capabilities. It has gained considerable attention in various fields. However, the randomly generated parameters in RVFL often lead to the loss of valid information and data redundancy, which severely degrades the model performance in practice. This paper first proposes an efficient network parameters learning approach for the original RVFL with pseudoinverse learner (RVFL-PL). Instead of taking the random feature mapping directly, RVFL-PL adopts a non-iterative manner to obtain influential enhancement nodes implanted with valuable information from input data, which realizes to improve the quality of the enhancement nodes and ease the problem caused by the randomly assigned parameters in the standard RVFL. Since the network parameters are optimized analytically, this improved variant can maintain the efficiency of the standard RVFL. Further, the RVFL-PL is extended to a multilayered structure (mRVFL-PL) to obtain high-level representations from the input data. The results of comprehensive experiments on some benchmarks indicate the performance improvement of the proposed method compared to other corresponding methods. [ABSTRACT FROM AUTHOR]

Process fault detection and diagnosis (FDD) is a predominant task to ensure product quality and process reliability in modern industrial systems. Those traditional FDD techniques are largely based on diagnostic experience. These methods have met significant challenges with immense expansion of plant scale and large numbers of process variables. Recently, deep learning has become the newest trends in process control. The upsurge of deep neural networks (DNNs) in leaning highly discriminative features from complicated process data has provided practitioners with effective process monitoring tools. This paper is to present a review and full developing route of deep learning-based FDD in complex process industries. Firstly, the nature of traditional data projection-based and machine learning-based FDD methods is discussed in process FDD. Secondly, the characteristics of deep learning and their applications in process FDD are illustrated. Thirdly, these typical deep learning techniques, e.g., transfer learning, generative adversarial network, capsule network, graph neural network, are presented for process FDD. These DNNs will effectively solve these problems of fault detection, fault classification, and fault isolation in process. Finally, the developing route of DNN-based process FDD techniques is highlighted for future work. [ABSTRACT FROM AUTHOR]

The proposed paper presents the analysis, design, implementation and evaluation of an ultra-short-term frequency trading system for the foreign exchange (FOREX) market, which features all stages of the trading process (Pretrade Analysis, Trend Forecasting, Transaction Execution) substantially exploiting artificial intelligence techniques. Our goal is to simulate the judgment and decision making of the human expert (technical analyst or broker) with a system that responds in a timely manner to changes in market conditions, thus allowing the optimization of ultra-short-term transactions. We designed and implemented a series of technical indicator simulators, which are fed to a novel artificial neural network architecture, to eventually generate the trend forecasting signal. We also designed and implemented a series of customizable ultra-short-term automated trading machines, which receive as inputs the generated forecasting signals and perform real-time virtual transactions. A comparative analysis of the results of both automated trading machines and each machine is carried out for a comprehensive variety of trend forecasting sources. [ABSTRACT FROM AUTHOR]

Taking the hot-rolling products made by an iron and steel company as the research object, this paper builds the inverse model reflecting the relationship between the hot-rolling steel product performance indicators, the chemical composition of steel and the rolling technological parameters by using the BP neural network. So the purpose of getting technological parameters is achieved, according to the given steel performance indicators. Combining the BP neural network, adaptive inverse control with internal model control theory, this paper builds the BP neural network inverse model with multiple input and single output based on internal model control. Therefore, it realizes the inverse mapping between the output and the input variables of the BP neural network. And the output variables can be obtained according to the input variables. Besides, this paper also gives the detailed steps to solve the inverse model. Then, the model is applied to the hot-rolling steel products quality control system. The performance indicators of the hot-rolling products are set up, and the rolling technological parameters—the rolling crimp temperature—are solved. The model realizes the controllability of rolling technological parameters. Finally, through the verification of hot-rolling products quality control positive system, the error is in line with the enterprise production requirements. [ABSTRACT FROM AUTHOR]

Permeability is an important parameter for oil and gas reservoir characterization. Permeability can be traditionally determined by well testing and core analysis. These conventional methods are very expensive and time-consuming. Permeability estimation in heterogeneous carbonate reservoirs is a challenge task to be handled accurately. Many researches tried to relate permeability and reservoir properties using complex mathematical equations which resulted in inaccurate estimation of the formation permeability values. Permeability prediction based on well logs using artificial intelligent techniques was presented by many authors. They used several wire-line logs such as gamma ray, neutron porosity, bulk density, resistivity, sonic, spontaneous potential, hole size, depths, and other logs. The objective of this paper is to develop an artificial neural network (ANN) model that can be used to predict the permeability of heterogeneous reservoir based on three logs only, namely resistivity, bulk density, and neutron porosity. In addition to the ANN model, in this paper and for the first time a mathematical equation from the ANN model will be extracted that can be used for permeability prediction for any data set without the need for the ANN model. Also, in this study and for the first time we introduced a new term which is the mobility index that can be used effectively in the permeability prediction. Mobility index term is derived from the mobile oil saturation that occurred due to the drilling fluid filtrate invasion. The obtained results showed that ANN model gave a comparable results with support vector machine and adaptive neuro-fuzzy inference system model. The developed mathematical equation from ANN model can be used to estimate the permeability for heterogamous carbonate reservoir based only on three parameters: bulk density, neutron porosity, and mobility index. Actual core data points (1223 points) with the three logs were used to train (857 data points, 70% of the data) and test the model for unseen data (366 data points, 30% of the data). The correlation coefficient for training and testing was 0.95, and the root-mean-square error was 0.28. The developed mathematical equation will help the engineers to save time and predict the permeability with a high accuracy using inexpensive technique. Introducing the new parameter, mobility index, in the prediction process greatly improved the permeability prediction from the log data compared to the actual measured data. [ABSTRACT FROM AUTHOR]

Water distribution system design is inherently associated with hydraulic calculations that require thorough evaluation of obtained results and accuracy of the applied solution. Currently, there are no programs that will replace a designer in these tasks, and there likely will not be such programs. However, some individuals are trying to develop computer programs featuring a certain degree of creativity to facilitate user decision making. In water distribution system design and hydraulic calculations, one should, inter alia, check pressure heads in different parts of the system. It is also important to establish whether the system should contain one or more pressure zones. This determination is connected with the appropriate location of booster and pressure reducing stations. In this paper, the nominal variable is defined. The classes of this variable describe problems related to a value of pressure and division of the water distribution system into pressure zones. By choosing one of the classes, an artificial neural network determines the problems that may arise in a given part of the water distribution system. The classification is conducted based on neural network input variables describing the specific parameters that affect water distribution system design, such as land relief, loss of pressure, pipe roughness and distance to a water supply. The paper presents a new approach that extends traditional methods of hydraulic calculations for water distribution systems by introducing the evaluation of a pressure head and the analysis of design concepts of pressure zones by using artificial neural networks. [ABSTRACT FROM AUTHOR]

This paper mainly studies the parameters estimation and synchronization of coupling recurrent dynamical neural networks (CRDNNs). Here, the weights and coupling parameters of CRDNNs are unknown. By designing a new adaptive controller and parameters update rules, a simple linear matrix inequality criterion is established for the parameters estimation and synchronization problem of CRDNNs. We prove the conclusion by constructing a suitable Lyapunov function. A numerical simulation is given to show the effectiveness of the conclusion. [ABSTRACT FROM AUTHOR]

This paper develops an online algorithm based on policy iteration for optimal control with infinite horizon cost for continuous-time nonlinear systems. In the present method, a discounted value function is employed, which is considered to be a more general case for optimal control problems. Meanwhile, without knowledge of the internal system dynamics, the algorithm can converge uniformly online to the optimal control, which is the solution of the modified Hamilton–Jacobi–Bellman equation. By means of two neural networks, the algorithm is able to find suitable approximations of both the optimal control and the optimal cost. The uniform convergence to the optimal control is shown, guaranteeing the stability of the nonlinear system. A simulation example is provided to illustrate the effectiveness and applicability of the present approach. [ABSTRACT FROM AUTHOR]

Reinforcement learning (RL), powered by artificial neural networks, has made considerable advances in agent autonomy, both in simulation-based and real-world environments. Millan-Arias et al. studied the behavior of RL agents in a proxemic-based environment, where the human comfort levels are dependent on the agent's proximity. Persiani and Hellstrom proposed a method to augment an RL agent with legibility, allowing the policy to be regularized after training, without having to modify the underlying RL algorithm. [Extracted from the article]

This paper is an addendum to Ref. Neural Computing and Applications (2021) 33:8981–9004 that adds information to clarify certain aspects of the models used adding information and some citations. In this paper an approach based on Co-active Neuro-Fuzzy Inference System named CANFIS-ART is proposed to automate data imputation procedure. This model is compared to other state-of-the-art imputation techniques such as its baseline model Artificial Neural Network and statistical methods, using a total of eighteen databases exposed to a perturbation procedure based on the random generation of non-monotone missing values pattern. A comparison of databases imputed by these models using a set of three classifiers were conducted. [ABSTRACT FROM AUTHOR]

Signature verification has been one of the major researched areas in the field of computer vision. Many financial and legal organizations use signature verification as an access control and authentication. Signature images are not rich in texture; however, they have much vital geometrical information. Through this work, we have proposed a signature verification methodology that is simple yet effective. The technique presented in this paper harnesses the geometrical features of a signature image like center, isolated points, connected components, etc. and, with the power of artificial neural network classifier, classifies the signature image based on their geometrical features. Publicly available dataset MCYT, BHSig260 (contains the image of two regional languages Bengali and Hindi) has been used in this paper to test the effectiveness of the proposed method. We have received a lower equal error rate on MCYT 100 dataset and higher accuracy on the BHSig260 dataset. [ABSTRACT FROM AUTHOR]

The introduction of logistics theory and logistics technology has made the government and enterprises gradually realize that the development of logistics has an important strategic role, which can effectively solve the changing needs of users, optimize resource allocation, improve the investment environment, and enhance the overall strength and overall competitiveness of the regional economy. This paper carries out matrix–vector multiplication operations and weight update operations, designs a perceptron neural network model, and realizes a simulation platform based on MLP neural network. Moreover, on the basis of the standard MLP neural network, this paper proposes to use the deep learning training mechanism to improve the MLP neural network, which provides effective technical support for the improvement of the prediction model. In addition, through the fusion of deep learning and MLP neural network, an MLP neural network with three hidden layers is determined. Finally, this paper builds a model based on the MLP neural network algorithm, selects the RBF kernel function as the kernel function of the model by referring to the relevant literature, and uses PSO to optimize the combination of parameters. It can be seen from the result of the evaluation index that each evaluation index is relatively small. The result shows that the prediction is accurate, and the empirical result shows the feasibility of the model to predict the demand for industrial logistics in Shanxi Province. [ABSTRACT FROM AUTHOR]

Recent years have witnessed the exponential growth of time series data as the popularity of sensing devices and development of IoT techniques; time series classification has been considered as one of the most challenging studies in time series data mining, attracting great interest over the last two decades. According to the empirical evidences, temporal representation learning-based time series classification has more superiority of accuracy, efficiency and interpretability as compared to hundreds of existing time series classification methods. However, due to the high time complexity of feature process, the performance of these methods has been severely restricted. In this paper, we first presented an efficient shapelet transformation method to improve the overall efficiency of time series classification, and then, we further developed a novel enhanced recurrent neural network model for deep representation learning to further improve the classification accuracy. Experimental results on typical real-world datasets have justified the superiority of our models over several shallow and deep representation learning competitors. [ABSTRACT FROM AUTHOR]

Misclassification is a critical problem in many machine learning applications. Since even the classifier models with high accuracy (e.g., > 95%) still introduce some misclassification error, it may not be possible to rely on the output of a classifier. In this paper, we introduce trustable learning, which prompts the learning model to yield only the true output, thus avoiding misclassifications. Whenever the model cannot decide the output accurately, the learning model should indicate that there could be a misclassification error if it is forced to classify, and hence, it should reject to make a decision or defer it to a human expert. Therefore, we develop a methodology for trustable learning and apply it to artificial neural networks and show that it is possible to develop a classifier with 0% misclassification error. We propose a novel neural network architecture named WisdomNet that could provide zero prediction error by introducing an additional neuron named as conjugate neuron that would indicate whether the network is able to classify the data correctly or not. The WisdomNet architecture can be applied to any previously built model, and we have evaluated WisdomNet with several network architectures such as multilayer perceptron, convolutional neural network, and deep network on different data sets. The results show that the WisdomNet is able to reduce the classification error rate to 0%, while labeling the data is difficult to classify as 'reject' at a low percentage of within around 10%. [ABSTRACT FROM AUTHOR]

The cost function gradient vanishing or exploding problem and slow convergence speed are key issues when training deep neural networks (DNNs). In this paper, we investigate the forward and backward propagation processes of DNN training and explore the properties of the activation function and derivative function (ADF) employed. The outputs' distribution of ADF with near-zero mean is proposed to reduce gradient problems. Additionally, the constant energy transfer of propagating data in the training process is also proposed to speed up convergence further. Based on wavelet frame theory, we derive a novel ADF, i.e., tight frame wavelet activation function (TFWAF) and tight frame wavelet derivative function (TFWDF) of the Mexican hat wavelet, to stabilize and accelerate DNN training. The nonlinearity of wavelet functions can strengthen the learning capacity of DNN models, while the sparse property of wavelets derived can reduce the overfitting problem and enhance the robustness of models. Experiments demonstrate that the proposed method stabilizes the DNN training process and accelerates convergence. [ABSTRACT FROM AUTHOR]

Manual calculation of the compressive strength of concrete (CSC) is an expensive and time-consuming process. Soft computing methods outperform the statistical methods used to resolve these problems. Nonetheless, complicated prediction models are still incomplete and require more exploration. Artificial neural networks (ANNs) provide a better and faster technique featuring solitary hidden layers and have improved the generalization capacity. The present paper presents three ANN-based (shuffled complex evolution, evaporation rate based water cycle algorithm (ERWCA), and Cuckoo optimization algorithm) prediction models to anticipate the compressive strength of concrete efficiently. An available database from the UCI repository is employed to develop and access the model performance. A comparison is made between the prediction accuracies of the above three techniques. Using all models, a comparative investigation has been conducted to predict the compressive strength of concrete at the curing ages of 91, 56, and 28 days. The experimental findings obtained from the ERWCA-MLP method indicate its capability of robust CSC prediction. On average, this method achieves the minimum RMSE of 0.55314 and 0.43329 and R2 of 0.99803 and 0.99824. The statistical significance test and the comparative analysis of simulation results indicate the superiority of ERWCA-MLP in predicting the compressive strength of concrete. [ABSTRACT FROM AUTHOR]

This paper introduces an effective method for improving dependency parsing which is based on a graph embedding model. The model helps extract local and global connectivity patterns between tokens. This method allows neural network models to perform better on dependency parsing benchmarks. We propose to incorporate node embeddings trained by a graph embedding algorithm into a bidirectional recurrent neural network scheme. The new model outperforms a baseline reference using a state-of-the-art method on three dependency treebanks for both low-resource and high-resource natural languages, namely Indonesian, Vietnamese and English. We also show that the popular pretraining technique of BERT would not pick up on the same kind of signal as graph embeddings. The new parser together with all trained models is made available under an open-source license, facilitating community engagement and advancement of natural language processing research for two low-resource languages with around 300 million users worldwide in total. [ABSTRACT FROM AUTHOR]

No-reference (NR) image quality assessment (IQA) is an important task of computer vision. Most NR-IQA methods via deep neural networks do not reach desirable IQA performance and have bulky models which make them difficult to be used in the practical scenarios. This paper proposes a lightweight transformer and multi-head prediction network for NR-IQA. The proposed method consists of two lightweight modules: feature extraction and multi-head prediction. The module of feature extraction exploits lightweight transformer blocks to learn features at different scales for measuring different image distortions. The module of multi-head prediction uses three weighted prediction blocks and an FC layer to aggregate the learned features for predicting image quality score. The weighted prediction block can measure the importance of different elements of input feature at the same scale. Since the importance of feature elements at the same scale and the importance of the features at different scales are both considered, the module of multi-head prediction can provide more accurate prediction results. Extensive experiments on the standard IQA datasets are conducted. The results show that the proposed method outperforms some baseline NR-IQA methods in IQA performance on the large image datasets. For the model complexity, the proposed method is also superior to several recent NR-IQA methods. [ABSTRACT FROM AUTHOR]

Memristive devices being applied in neuromorphic computing are envisioned to significantly improve the power consumption and speed of future computing platforms. The materials used to fabricate such devices will play a significant role in their viability. Graphene is a promising material, with superb electrical properties and the ability to be produced in large volumes. In this paper, we demonstrate that a graphene-based memristive device could potentially be used as synapses within spiking neural networks (SNNs) to realise spike timing-dependant plasticity for unsupervised learning in an efficient manner. Specifically, we verify the operation of two SNN architectures tasked for single-digit (0–9) classification: (i) a single layer network, where inputs are presented in 5 × 5 pixel resolution, and (ii) a larger network capable of classifying the dataset. Our work presents the first investigation and large-scale simulation of the use of graphene memristive devices to perform a complex pattern classification task. In favour of reproducible research, we will make our code and data publicly available. This can pave the way for future research in using graphene devices with memristive capabilities in neuromorphic computing architectures. [ABSTRACT FROM AUTHOR]

The convolutional neural network has been proven to be a robust recognition and diagnosis model for modelling diseases. In general, a convolutional neural network consists of feature learning which is considered a vital process in terms of feature extraction and classifier algorithms used for decision-making. In recent years, various pooling operations considered the spine of the feature learning layer have been developed in the literature; unfortunately, few of them can be spread to tackle the overfitting problem. In this paper, a novel pooling approach termed whole image average pooling is proposed to hold over the most significant feature information on the learning layer. The proposed method consists of two main stages: in the first stage, the average of all the values of the input image is calculated, in the second stage, the input image is divided into regions that represent the feature set of the input image, and then, the values of the regions are calculated. While obtaining the values of these features, the average of all values of the input image is taken into account. The proposed whole image average pooling approach addresses the overfitting problem by processing all pixel values in the image data and is used to get the important feature information from the medical datasets. To evaluate the performance of the proposed whole image average pooling, four clinically applicable tumour datasets as medical datasets have been trained and tested on seven convolutional neural network models AlexNet, VGG16, ResNet18, GoogleNet, ResNet50, R-CNN, and Faster R-CNN. The experimental outcomes demonstrate that the proposed whole image average pooling consistently outperformed the state-of-the-art pooling operation in terms of accuracy, sensitivity, specificity, positive predictive value, false-positive rate, and F1-measure evaluation criteria in three datasets. [ABSTRACT FROM AUTHOR]

Deep neural networks are known to be vulnerable to various types of adversarial attacks, especially word-level attacks, in the field of natural language processing. In recent years, various defense methods are proposed against word-level attacks; however, most of those defense methods only focus on synonyms substitution-based attacks, while word-level attacks are not based on synonym substitution. In this paper, we propose a textual adversarial defense method against word-level adversarial attacks via textual embedding based on the semantic associative field. More specifically, we analyze the reasons why humans can read and understand textual adversarial examples and observe two crucial points: (1) There must be a relation between the original word and the perturbed word or token. (2) Such a kind of relation enables humans to infer original words, while humans have the ability to associations. Motivated by this, we introduce the concept of semantic associative field and propose a new defense method by building a robust word embedding, that is, we calculate the word vector by exerting the related word vector to it with potential function and weighted embedding sampling for simulating the semantic influence between words in same semantic field. We conduct comprehensive experiments and demonstrate that the models using the proposed method can achieve higher accuracy than the baseline defense methods under various adversarial attacks or original testing sets. Moreover, the proposed method is more universal, while it is irrelevant to model structure and will not affect the efficiency of training. [ABSTRACT FROM AUTHOR]

Due to the noises captured in ultrasound device and image reconstruction process, the edges of thyroid nodule are usually not distinctive and it is very difficult for existing approaches to well segment them in ultrasound images. While deep neural networks like U-Net have been successfully applied in many medical image segmentation tasks, their segmentation performances on ultrasound images are still not satisfactory. To address this issue, we propose in this paper a boundary field regression branch to provide useful boundary information to help improve the segmentation performance of existing networks. Without requirement of additional labeling costs, our approach firstly generates boundary field heatmap from available segmentation masks, which are then used as a supervision to train the regression branch. As a general architecture, our branch can be integrated with all encoder-decoder like segmentation networks. A dataset consisting of 3169 images from 2004 patients is used for experiments. We integrate our branch with U-Net, Attention U-Net, U-Net++ and DeepLabv3+; consistent improvements of Dice metrics were observed. The memory and computation costs required by adding our branch are marginal as well. [ABSTRACT FROM AUTHOR]

Despite the enormous performance of deep neural networks (DNNs), recent studies have shown their vulnerability to adversarial examples (AEs), i.e., carefully perturbed inputs designed to fool the targeted DNN. Currently, the literature is rich with many effective attacks to craft such AEs. Meanwhile, many defense strategies have been developed to mitigate this vulnerability. However, these latter showed their effectiveness against specific attacks and does not generalize well to different attacks. In this paper, we propose a framework for defending DNN classifier against adversarial samples. The proposed method is based on a two-stage framework involving a separate detector and a denoising block. The detector aims to detect AEs by characterizing them through the use of natural scene statistic (NSS), where we demonstrate that these statistical features are altered by the presence of adversarial perturbations. The denoiser is based on block matching 3D (BM3D) filter fed by an optimum threshold value estimated by a convolutional neural network (CNN) to project back the samples detected as AEs into their data manifold. We conducted a complete evaluation on three standard datasets, namely MNIST, CIFAR-10 and Tiny-ImageNet. The experimental results show that the proposed defense method outperforms the state-of-the-art defense techniques by improving the robustness against a set of attacks under black-box, gray-box and white-box settings. The source code is available at: https://github.com/kherchouche-anouar/2DAE. [ABSTRACT FROM AUTHOR]

Age estimation from face images can be profitably employed in several applications, ranging from digital signage to social robotics, from business intelligence to access control. Only in recent years, the advent of deep learning allowed for the design of extremely accurate methods based on convolutional neural networks (CNNs) that achieve a remarkable performance in various face analysis tasks. However, these networks are not always applicable in real scenarios, due to both time and resource constraints that the most accurate approaches often do not meet. Moreover, in case of age estimation, there is the lack of a large and reliably annotated dataset for training deep neural networks. Within this context, we propose in this paper an effective training procedure of CNNs for age estimation based on knowledge distillation, able to allow smaller and simpler "student" models to be trained to match the predictions of a larger "teacher" model. We experimentally show that such student models are able to almost reach the performance of the teacher, obtaining high accuracy over the LFW+, LAP 2016 and Adience datasets, but being up to 15 times faster. Furthermore, we evaluate the performance of the student models in the presence of image corruptions, and we demonstrate that some of them are even more resilient to these corruptions than the teacher model. [ABSTRACT FROM AUTHOR]

In this paper, a new fine-grained image classification (FGIC) network with feature relationship enhancement of multiple stages is established. After the engaging of scene text in FGIC and retrieval, basic architecture of local, global, text feature encoders and classifier have been approved. This method retains these portions and expands them into a five-module architecture. In specific, positional encoding is incorporated to both local and textual feature encoders such that complementary information carried could engage in feature representation. In local and textual feature encoders, intra-modal semantic relation reasoning is introduced for FGIC by a proposed General Feature Relation Enhancement (GFRE) module. GFRE is a feature reasoning module applicable to any two inputs of same modality or distinct modalities. GFRE adopts Graph Attention which represents and infers relationships among graph data. Moreover, latest multi-modal reasoning module is improved by a proposed Multi-Head Multi-Modal Joint Semantic Reasoning module consisted of cross-modal GFREs by multi-head fusion. Experimental results on multiple datasets verify the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

The application of clustering in generating random subspace has improved the accuracy and diversity of ensemble classification methods. If clusters are not balanced (unequal size of clusters) and not strong (unequal number of data from each class in each cluster), the results will deviate from classes with more samples in each cluster and thereby will be biased. The current paper presents a novel strong balance-constrained clustering or hard-strong clustering. This method creates diverse strong balanced data clusters to train different base classifiers and an artificial neural network with more than one hidden layer, in which the final decision is made for the data class through majority voting. By implementing the proposed method on 16 datasets, two objectives are followed: enhancing the performance of ensemble classifier and deep learning-based method (data mining objective), and adopting appropriate policies for budget, time, and energy assignment to various business domains by decision-makers (business objective). Based on the evaluation and comparison of the results, the proposed method is faster than other balancing methods. Furthermore, the accuracy of the proposed ensemble method has proved acceptable improvement than other ensemble classification methods. [ABSTRACT FROM AUTHOR]

Adversarial examples can attack multiple unknown convolutional neural networks (CNNs) due to adversarial transferability, which reveals the vulnerability of CNNs and facilitates the development of adversarial attacks. However, most of the existing adversarial attack methods possess a limited transferability on vision transformers (ViTs). In this paper, we propose a partial blocks search attack (PBSA) method to generate adversarial examples on ViTs, which significantly enhance transferability. Instead of directly employing the same strategy for all encoder blocks on ViTs, we divide encoder blocks into two categories by introducing the block weight score and exploit distinct strategies to process them. In addition, we optimize the generation of perturbations by regularizing the self-attention feature maps and creating an ensemble of partial blocks. Finally, perturbations are adjusted by an adaptive weight to disturb the most effective pixels of original images. Extensive experiments on the ImageNet dataset are conducted to demonstrate the validity and effectiveness of the proposed PBSA. The experimental results reveal the superiority of the proposed PBSA to state-of-the-art attack methods on both ViTs and CNNs. Furthermore, PBSA can be flexibly combined with existing methods, which significantly enhances the transferability of adversarial examples. [ABSTRACT FROM AUTHOR]

This paper introduces a developed method of an artificial neural networks (ANN) architecture for estimating the residual strength of clay soil. To implement this purpose, a database of input soil parameters is built, including liquid limit, plasticity index, A-line value, clay fraction, massive minerals, mica, kaolinite, and smectite, in which the output is the residual friction angle. The ANN model was developed by extensively analyzing a number of hidden layers and number of neurons in every layer, incorporating a statistical investigation of the model performance. The obtained results indicate that the ANN model is an outperformed and promising method based on various well-known indicators such as correlation coefficient, mean absolute error, and root mean square error. The achieved ANN model also gives higher estimation accuracy than those results in the literature. Finally, partial dependence plot 2-D was used for sensitivity analysis within the ANN algorithm to investigate the effect of coupled input variables on the estimated residual friction angle of the soil. It was found that A-line value, clay fraction, and massive minerals are the most important input parameters influencing the residual friction angle. [ABSTRACT FROM AUTHOR]

In the previous works on heat stress detection, various state-of-the-art machine learning techniques had been used to detect stress patterns from electroencephalographic (EEG) signals. Since the handcrafted feature engineering-based approaches pose certain limitations and are sensitive to transform the nonlinearities of EEG, deep learning techniques have drawn attention in the domain of EEG-based applications. Moreover, existing approaches for stress detection consider the whole frequency band (delta to gamma) which conceals the redundant and lossy information that increased the false detection rate. Also, these approaches were implemented only to identify the stress in binary classes and confined their application to identifying the level of stress. Therefore, a multi-scale channel-wise convolution-based multi-level heat stress assessment has been proposed in this paper. The novel contributions of this work are (1) designing a multi-scale convolutional neural network (MS-CNN) for extracting precise information from individual frequency bands of EEG, (2) use of sparse connectivity to reduce the redundant information and increase the performance with less trainable parameters, and (3) multi-level heat stress assessment for precise identification of the severity of stress. The proposed approaches were evaluated on pre-recorded data of 10 rats in a simulated laboratory environment. The high accuracy of approximately 96–97% and 90–92% has been achieved for binary and multi-level stress detection. There is approximately a 6 to 50% reduction in the trainable parameters with a 2% improvement of accuracy achieved on adopting channel-wise convolution over MS-CNN and deep convolutional neural network (DCNN). This demonstrates the effectiveness of the adopted multiscale feature extraction and sparse connectivity to improve the performance with a less complex model. [ABSTRACT FROM AUTHOR]

Diabetic Retinopathy (DR), one of the most progressive sight-threatening diseases caused by the long-term diabetic condition, can lead to vision impairment and blindness later. Early diagnosis and timely treatment help control and avert DR from its progression. However, manual grading is exceptionally challenging and arduous due to the complex anatomical features in the retina. Therefore, developing an automated diagnostic method for screening DR is obligatory. This paper proposes a deep learning-enabled optimized feature selection approach to classify the stage of DR severity in the fundus image. At first, a pre-processing phase eradicates the noise and improves the contrast in the retinal fundus image. Then, blood vessel segmentation is performed using the Coherence Enhancing Energy Based Regularized Level Set Evolution method in the green channel fundus image. Subsequently, the optic disk is segmented with Canny Anisotropic Diffusion filter and morphological transformations. Next, the candidate lesion region is detected using an Attention-based Fusion Network (AFU-Net). Then, shape and texture features are extracted, and then, the optimal subset of features is selected using the Improved Harris Hawk Optimization algorithm. Finally, a deep Convolutional Neural Network classifies the DR stages, and the model weight is updated using the same algorithm. The proposed method achieved superior performance in two benchmark public datasets compared with the existing state-of-the-art methods using F1-score, accuracy, sensitivity, and specificity measures. [ABSTRACT FROM AUTHOR]

Advances in deep neural networks have led to significant improvement of object detection accuracy. However, object detection in crowded scenarios is a challenging task for neural networks since extremely overlapped objects provide fewer visible cues for a model to learn from. Further complicating the detection of overlapping objects is the fact that most object detectors produce multiple redundant detections for single objects, which are indistinguishable from detections of separate overlapped objects. Most existing works use some variant of non-maximum suppression to prune duplicate candidate bounding boxes based on their confidence scores and the amount of overlap between predicted bounding boxes. These methods are unaware of how much overlap there actually is between the objects in the image, and are therefore inclined to merge detections for highly overlapped objects. In this paper, we propose an overlap aware box selection solution that uses a predicted overlap map to help it decide which highly overlapping bounding boxes are associated with actual overlapping objects and should not be pruned. We show our solution outperforms the state-of-the-art set-NMS bounding box selection algorithm for both the crowdHuman dataset and a sports dataset. [ABSTRACT FROM AUTHOR]

Object tracking has achieved impressive performance in computer vision. However, there are many challenges due to complex scenarios in reality. The mainstream trackers mostly locate the object in form of two branches, which limits the ability of trackers to fully mine similarity between template and search region. In this paper, we propose a multi-branch and multi-scale perception object tracking framework based on Siamese Convolutional Neural Networks (MultiBSP), in which the multi-branch tracking framework is established based on the idea of relation mining, and a tower-structured relation network is designed for each branch to learn the non-linear relation function between template and search region. By branch combination, multiple branches can verify their predictions with each other, which is beneficial to robust tracking. Besides, in order to sense the scale and aspect ratio of object in advance, a multi-scale perception module is designed by utilizing the dilated convolutions in five scales, which contributes to the ability of tracker to deal with scale variation. In addition, we propose an information enhancement module that focuses on important features and suppresses unnecessary ones along spatial and channel dimensions. Extensive experiments on six visual tracking benchmarks including OTB100, VOT2018, VOT2019, UAV123, GOT-10k, and LaSOT demonstrate that our MultiBSP can achieve robust tracking and have state-of-the-art performance. Finally, ablation experiments verify the effectiveness of each module and the tracking stability is proved by qualitative and quantitative analyses. [ABSTRACT FROM AUTHOR]

The increase in debt levels of families in different parts of the world has drawn the attention of organizations dedicated to the prevention of financial risk and has highlighted the need to develop early detection methods for over-indebtedness. In this paper, we propose a hybrid model of the adaptive neural fuzzy inference system (ANFIS) and Probit model for the prediction of household over-indebtedness. The proposed model is compared with Probit, artificial neural networks (ANN), classification and regression trees (CART), random forest (RF) and support vector machine (SVM) models. The most relevant parameters for the performance of each model are optimized, and we address data balance problems through the synthetic minority over-sampling technique (SMOTE). We use data obtained from the Financial Household Survey of the Central Bank of Chile. The results show that the proposed model performs significantly better than the reference models in terms of the correct classification of indebted individuals. Consequently, this model provides an innovative understanding of household over-indebtedness, which can be useful for different governmental entities focused on preventing excessive indebtedness and maintaining financial stability. [ABSTRACT FROM AUTHOR]

Fused deposition modeling (FDM) has shown to be a highly beneficial process for product development. However, one of the great challenges in using FDM is maintaining the surface quality of the produced part. Poor texture quality can be regarded as a defect. It is not part of the geometric prototype but results from the fabrication process. Poor input parameters typically cause these defects by the user. This paper presents the integration between an artificial neural network (ANN) and symbiotic organism search, known as ANN–SOS, to model and minimize the surface roughness (Ra) of the FDM process. The FDM input parameters considered were layer height, print speed, print temperature, and outer shell speed. The experimental data were collected using the central composite design response surface method. Then, the surface roughness model was established using an ANN. After validating the model's accuracy, it was combined with symbiotic organism search (SOS) to determine the optimal parameter settings for the minimum surface roughness value. The results illustrate that ANN–SOS with a 4-8-8-1 network structure would be the best model for surface roughness prediction. It was observed that decreasing the layer thickness, printing speed, print temperature, and outer shell speed of the FDM input parameters for ANN–SOS resulted in minimum surface roughness of approximately 2.011 µm, which was 12.36% better than the RSM method. [ABSTRACT FROM AUTHOR]

Similarity searches on heterogeneous information networks (HINs) have attracted wide attention from both industrial and academic areas in recent years; for example, they have been used for friend detection in social networks and collaborator recommendation in coauthor networks. The structural information on the HIN can be captured by multiple metapaths, and people usually utilize metapaths to design methods for similarity search. The rich semantics in HINs are not only structural information but also content stored in nodes. However, the content similarity of nodes was usually not valued in the existing methods. Although some researchers have recently considered both types of information in machine learning-based methods for similarity search, they have used structure and content information separately. To address this issue by balancing the influence of structure and content information flexibly in the process of searching, we propose a double channel convolutional neural network model for top-k similarity search, which uses path instances as model inputs and generates structure and content embeddings for nodes based on different metapaths. We design an attention mechanism to enhance the differences in metapaths for each node. Another attention mechanism is used to combine the content and structure information of nodes. Finally, an importance evaluation function is designed to improve the accuracy and make the model more explainable. The experimental results show that our search algorithm can effectively support top-k similarity search in HINs and achieve higher performance than existing approaches. [ABSTRACT FROM AUTHOR]

Pedestrian Attribute Recognition (PAR) can provide valuable clues for several innovative surveillance applications. It is also a difficult task because inference of the multiple attributes at a far distance is challenging in real complex scenarios. Most existing methods improve the PAR with visual attention mechanisms or body-part detection modules, which increase the complexity of networks and require manual annotations of the human body. Also, uneven data distribution, leading to a decline in recall values, is still underestimated. This paper presents a novel multi-label optimization algorithm to mitigate these issues, named Multi-label Contrastive Focal Loss (MCFL). Specifically, we first propose a multi-label focal loss to emphasize the error-prone and minority attributes with a separated re-weighting scheme. And then, we introduce a multi-label contrastive learning strategy based on the multi-label divergences to help the deep network to distinguish the hard fine-grained attributes. We conduct extensive experiments on seven PAR benchmarks, and results indicate that the proposed MCFL with the native ResNet-50 backbone surpasses the state-of-the-art comparison methods in mean accuracy and recall. [ABSTRACT FROM AUTHOR]

Artificial neural network (ANN) is one of the technologies used for emerging real-world problems. Activation functions (AF) are used in deep learning architectures to make decisions in the hidden and output layers. An AF influences the dynamics of training and performance of an ANN. This paper proposes a novel AF called E-Tanh by extending the Tanh AF for ANN models. When used in three shallow ANN models with 2, 4, and 6 hidden layers with the Modified National Institute of Standards and Technology database (MNIST), it gave a maximum accuracy of 99.5% for handwritten digits classification than the results produced by the existing AFs. The proposed AF E-Tanh outperformed other existing and well-known AFs ReLU, Sigmoid, Tanh, Swish, and E-Swish. In deep ANNs with 5–50 hidden layers, the proposed AF's prediction accuracy of up to 25 hidden layers was higher than the existing AFs on the MNIST database. When used in shallow CNN for the CIFAR10 database, it outperformed other AFs. When used in wide residual network, it performed poorly in classifying images in CIFAR10 and CIFAR100 database but performed competitively with all other AFs in CNN models on the MNIST dataset. [ABSTRACT FROM AUTHOR]

Recently, many studies have been carried out on model compression to handle the high computational cost and high memory footprint brought by the implementation of deep neural networks. In this paper, model compression of convolutional neural networks is constructed as a multiobjective optimization problem with two conflicting objectives, reducing the model size and improving the performance. A novel structured pruning method called Conventional-based and Evolutionary Approaches Guided Multiobjective Pruning (CEA-MOP) is proposed to address this problem, where the power of conventional pruning methods is effectively exploited for the evolutionary process. A delicate balance in pruning rate and model accuracy has been automated achieved by a multiobjective optimization evolutionary model. First, an ensemble framework integrates pruning metrics to establish a codebook for further evolutionary operations. Then, an efficient coding method is developed to shorten the length of chromosome, thus ensuring its superior scalability. Finally, sensitivity analysis is automatically carried out to determine the upper bound of pruning rate for each layer. Notably, on CIFAR-10, CEA-MOP reduces more than 50% FLOPs on ResNet-110 and improves the relative accuracy. Moreover, on ImageNet, CEA-MOP reduces more than 50% FLOPs on ResNet-101 with negligible top-1 accuracy drop. [ABSTRACT FROM AUTHOR]

This paper concerns the global exponential synchronization of coupled neural networks with stochastic perturbations and mixed time-varying delays. To be more practical, we assume that the communication topology arbitrarily switches among a finite set of directed topologies, each of which is only required to have a directed spanning tree. Moreover, we assume that there are impulsive effects in the process of signal exchanging. We will show that all the stochastic dynamical neural networks can achieve exponential synchronization even if only a single impulsive controller is exerted. Some sufficient synchronization criteria are given based on multiple Lyapunov theory. A simple example is presented to show the application of the criteria obtained in this paper. [ABSTRACT FROM AUTHOR]

Under specific terrain and climatic conditions, it is extremely easy to cause various types of geological hazards, and the occurrence of geological hazards will affect people's production and life, with the consequence that the economic losses are extremely great, and even severely endanger human life. However, the existing geological hazard danger zoning is slightly insufficient in accuracy and operating efficiency, and the effect in practical application needs to be further improved. In view of the above problems, this paper proposes a study on the application of GIS-based artificial neural network models in the geological disaster risk zoning. This article first expounds the related concepts of geological hazards zoning and gives the principles to be followed. Then, the calculation of evaluation factors and risk level probabilities are proposed, and a hierarchical model is constructed using the analytic hierarchy process. The weight of each evaluation factor is multiplied with the information to obtain the weighted information. Calculate the probability based on the topographic features, stratigraphic lithology, and terrain slope. Combine the artificial neural network with BP neural network to predict the results. Through the simulation experiments on the geological data of the low-mountain and hilly areas in Wanli District, Nanchang, the results show that the proportion of high-probability-prone areas and medium-probability-prone areas predicted by this paper is as high as 91.87%, and the evaluation results are consistent with the actual disaster occurrence better. The area under the ROC curve was 87.3%, which also verified the effectiveness of the method in this paper. [ABSTRACT FROM AUTHOR]

Groundwater flow in the Grasberg open-pit mine is governed by fractured flow media. Groundwater modeling in fractured media requires detailed hydraulic conductivity (K) value distribution to illustrate hydrogeological conditions of the Grasberg open-pit mine properly. The value of K is estimated by using the hydraulic conductivity (HC) system method based on rock quality designation (RQD), lithology permeability index (LPI), depth index (DI), and gouge content designation (GCD) data. Accordingly, all parameters must be available, but in this case, this information are only partially available. This paper proposes a method to solve this problem with artificial neural network (ANN) through a five approach scenario to find the K value with a model of incomplete empirical parameters. The artificial neural network back propagation (ANNBP) method is used to estimate the distributed K value based on the distributed observed RQD and LPI, as shown in the block model. This paper uses five optimum architectural schemes consisting of learning rate selection, momentum coefficient, number of nodes, number of hidden layers, activation method (sigmoid, tangent hyperbolic and Gaussian). Verification of the results of the K value is approached by the statistical approaches determination coefficient (R2), correlation coefficient (R), standard error of estimate (SEE), root-mean-squared (RMS), and normal root mean square (NRMS). ANNBP modeling used data for training and testing based on packer test and slug test as many as 49 points. Through five scenarios, it was found that scenario with learning rates of 10–5, momentum coefficient 0.1, number of nodes 10, number of hidden layer 5, with hyperbolic tangent activation methods is the most optimum result with R2 = 0.822, the accuracy of predicting/validating results is expressed by using R2 = 0.863, with R = 0.929, and the error percentage = 3.67%. The K value prediction results are used as an input for groundwater modeling. The modeling results show a significant correlation with high validity between the data extracted from the model and field observation. In a 3D model, the K value distribution is similar to the field RQD data distribution. Furthermore, the K value distribution results are clustered in order to understand the relationship between geological as well as geotechnical data. The highest K values are found in volcanic breccia rocks (DLMVB), volcanic sediments (DLMVS), volcanic andesite (DLMVA) and quaternary glacial till. The geological condition is confirmed by the average rock RQD value between 12.4 and 39.6%. [ABSTRACT FROM AUTHOR]

Simulation optimization studies the problem of optimizing simulation-based objectives. This field has a strong history in engineering but often suffers from several difficulties including being time-consuming and NP-hardness. Simulation optimization is a new and hot topic in the field of system simulation and operational research. This paper presents a hybrid approach that combines Evolutionary Algorithms with neural networks (NNs) for solving simulation optimization problems. In this hybrid approach, we use NNs to replace the known simulation model for evaluating subsequent iterative solutions. Further, we apply the dynamic structure-based neural networks to learn and replace the known simulation model. The determination of dynamic structure-based neural networks is the kernel of this paper. The final experimental results demonstrated that the proposed approach can find optimal or close-to-optimal solutions and is superior to other recent algorithms in simulation optimization. [ABSTRACT FROM AUTHOR]

It is a common practice to adjust the number of hidden neurons in training, and the removal of neurons in neural networks plays an indispensable role in this architecture manipulation. In this paper, a succinct and unified mathematical form is upgraded to the generic case for removing neurons based on orthogonal projection and crosswise propagation in a feedforward layer with different architectures of neural networks, and further developed for several neural networks with different architectures. For a trained neural network, the method is divided into three stages. In the first stage, the output vectors of the feedforward observation layer are classified to clusters. In the second stage, the orthogonal projection is performed to locate a neuron whose output vector can be approximated by the other output vectors in the same cluster with the least information loss. In the third stage, the previous located neuron is removed and the crosswise propagation is implemented in each cluster. On accomplishment of the three stages, the neural network with the pruned architecture is retrained. If the number of clusters is one, the method is degenerated into its special case with only one neuron being removed. The applications to different architectures of neural networks with an extension to the support vector machine are exemplified. The methodology supports in theory large-scale applications of neural networks in the real world. In addition, with minor modifications, the unified method is instructive in pruning other networks as far as they have similar network structure to the ones in this paper. It is concluded that the unified pruning method in this paper equips us an effective and powerful tool to simplify the architecture in neural networks. [ABSTRACT FROM AUTHOR]

This paper deals with modelling ship speed and fuel consumption using artificial neural network (ANN) techniques. These tools allowed us to develop ANN models that can be used for predicting both the fuel consumption and the travel time to the destination for commanded outputs (the ship driveline shaft speed and the propeller pitch) selected by the ship operator. In these cases, due to variable environmental conditions, making decisions regarding setting the proper commanded outputs to is extraordinarily difficult. To support such decisions, we have developed a decision support system. Its main elements are the ANN models enabling ship fuel consumption and speed prediction. To collect data needed for building ANN models, sea trials were conducted. In this paper, the decision support system concept, input and variables of the ship driveline system models, and data acquisition methods are presented. Based on them, we developed appropriate ANN models. Subsequently, we performed a quality assessment of the collected data set, data normalization and division of the data set, selection of an ANN model architecture and assessment of their quality. [ABSTRACT FROM AUTHOR]