Search

Although knowledge of the microstructure of food of vegetal origin helps us to understand the behavior of food materials, the variability in the microstructural elements complicates this analysis. In this regard, the construction of learning models that represent the actual microstructures of the tissue is important to extract relevant information and advance in the comprehension of such behavior. Consequently, the objective of this research is to compare two machine learning techniques—Convolutional Neural Networks (CNN) and Radial Basis Neural Networks (RBNN)—when used to enhance its microstructural analysis. Two main contributions can be highlighted from this research. First, a method is proposed to automatically analyze the microstructural elements of vegetal tissue, and second, a comparison was conducted to select a classifier to discriminate between tissue structures. For the comparison, a database of microstructural elements images was obtained from pumpkin (Cucurbita pepo L.) micrographs. Two classifiers were implemented using CNN and RBNN, and statistical performance metrics were computed using a 5-fold cross-validation scheme. This process was repeated one hundred times with a random selection of images in each repetition. The comparison showed that the classifiers based on CNN produced a better fit, obtaining F1–score average of 89.42% in front of 83.83% for RBNN. In this study, the performance of classifiers based on CNN was significantly higher compared to those based on RBNN in the discrimination of microstructural elements of vegetable foods.

This article describes the process of developing a prototype for anger detection by processing and analyzing electroencephalogram (EEG) and facial expression imaging. The prototype was developed in Python and is divided into functions which allow the creation of classification models for EEG bands and facial expressions, the prediction of previously unclassified file classes for each of the data inputs and the recording of video. The results obtained during the execution of the proof of concept, it was found that in the test data some sensors are better for the classification of characteristics since they obtained the highest averages in terms of efficiency. In the case of the EEG, two proofs of concept were carried out, the first one evaluated and classified the information from a five-second signal segment and it was found that the FP1, FP2 and F3 sensors were the best classified, obtaining a hit rate that ranged from 60% to 73%. In the second, the bands were evaluated, and it was found that Beta and Theta had a better hit rate in relation to other bands. With the data obtained, it was revealed that there is an average efficiency rate of 67% in the Beta band and 74% in the Theta band in the case of EEG, and a rate of 86% in the case of facial expressions.

The classification of fresh fruits according to their visual ripeness is typically a subjective and tedious task; consequently, there is a growing interest in the use of non-contact techniques to automate this process. Machine learning techniques, such as artificial neural networks, support vector machines (SVMs), decision trees, and K-nearest neighbor algorithms, have been successfully applied for classification problems in the literature, particularly for images of fruit. However, the particularities of each classification problem make it difficult, if not impossible, to select a general technique that is applicable to all types of fruit. In this paper, the combinations of four machine learning techniques and three color spaces (RGB, HSV, and L*a*b*) were evaluated with regard to their ability to classify Cape gooseberry fruits. To this end, 925 Cape gooseberry fruit samples were collected, and each fruit was manually classified into one of seven different classes according to its level of ripeness. The color values of each fruit image in the three color spaces and their corresponding ripening stages were organized for training and validation following a fivefold cross-validation strategy in an iterative process repeated 100 times. According to the results, the classification of Cape gooseberry fruits by their ripeness level was sensitive to both the color space and the classification technique used. The models based on the L*a*b* color space and the SVM classifier showed the highest f-measure regardless of the color space, and the principal component analysis combination of color spaces improved the performance of the models at the expense of increased complexity.

Software testing is an essential activity to ensure the quality of software systems. Combinatorial testing is a method that facilitates the software testing process; it is based on an empirical evidence where almost all faults in a software component are due to the interaction of very few parameters. The test generation problem for combinatorial testing can be represented as the construction of a matrix that has certain properties; typically this matrix is a covering array. Covering arrays have a small number of tests, in comparison with an exhaustive approach, and provide a level of interaction coverage among the parameters involved. This paper presents a repository that contains binary covering arrays involving many levels of interaction. Also, it discusses the importance of covering array repositories in the construction of better covering arrays. In most of the cases, the size of the covering arrays included in the repository reported here are the best upper bounds known, moreover, the files containing the matrices of these covering arrays are available to be downloaded. The final purpose of our Binary Covering Arrays Repository (BCAR) is to provide software testing practitioners the best-known binary test-suites.

Covering arrays are combinatorial designs used as test-suites in software and hardware testing. Because of their practical applications, the construction of covering arrays with a smaller number of rows is desirable. In this work we develop a greedy-metaheuristic 3-stage approach to construct covering arrays that improve some of the best-known ones. In the first stage, a covering perfect hash family is created using a metaheuristic approach; this initial array may not be complete, and so the derived covering array may have missing tuples. In the second stage, the covering perfect hash family is converted to a covering array and, in case there are missing tuples, a greedy approach completes the covering array through the addition of some rows. The third stage is an iterative postoptimization stage that combines two greedy algorithms and a metaheuristic algorithm; the greedy algorithms detect and reduce redundancy in the covering array, and the metaheuristic algorithm covers the tuples that may become uncovered after the reduction of redundancy. The effectiveness of our greedy-metaheuristic 3-stage approach is assessed through the construction of covering arrays of order four and strengths 3–6; the main results are the improvement of 9473 covering arrays of strength three, 9303 of strength four, 2150 of strength five, and 291 of strength six. To see how to apply covering arrays to real testing scenarios, the final part of this work presents the use of covering arrays of order four for setting up a composting process.

The evaluation of cheese texture during the ripening phase usually involves invasive and destructive methods, as well as specialized equipment, which are non-advantageous characteristics for routine tests. Therefore, new noninvasive technologies for measuring texture properties are being studied. In this paper, forty Swiss-type cheese samples were prepared and carried to the ripening stage. During this process, hyperspectral images (HSI) were obtained in reflectance mode, in a range of 400–1000 nm. The hardness of Swiss-type cheese was measured using the technique of texture profile analysis. The relationship between spectral profiles and hardness values was modeled using two types of regression models, i.e., partial least squares regression (PLSR) and artificial neural networks (ANN). For both PLSR and ANN, two models were created, the first one uses all the wavelengths and the second makes a selection of the relevant wavelengths. The ANN models showed slightly better performance than the PLSR models. As result, it is possible to use the proposed technique (HSI + ANN) to predict the texture properties of Swiss-type cheeses throughout the ripening period.

The use of machine learning techniques to automate the sorting of Cape gooseberry fruits according to their visual ripeness has been reported to provide accurate classification results. Classifiers like artificial neural networks, support vector machines, decision trees, and nearest neighbors are commonly employed to discriminate fruit samples represented in different color spaces (e.g., RGB, HSV, and L*a*b*). Although these feature spaces are equivalent up to a transformation, some of them facilitate classification. In a previous work, authors showed that combining the three-color spaces through principal component analysis enhances classification performance at expenses of increased computational complexity. In this paper, two combination and two selection approaches are explored to find the best characteristics among the combination of the different color spaces (9 features in total). Experimental results reveal that selection and combination of color channels allow classifiers to reach similar levels of accuracy, but combination methods require increased computational complexity.

The microstructure is a factor in the knowledge and prediction of properties in food and the associated changes during processing. The objective of this work was to evaluate the feasibility of using a convolution neural network (CNN) for the discrimination of structures in foods of vegetable origin. Micrographs of pumpkin were processed digitally to improve the detection of structures (cells and intercellular spaces). Later the found elements were classified in two sets, using a trained operator. The implementation made use of a pre-trained network AlexNet, performing cross-validation, and one hundred repetitions randomizing the information delivered to the training and validation processes. The statistics obtained were accuracy and F-measure. Therefore, the use of convolutional neural networks shows potential for the discrimination of structures in foods of vegetal origin.

Computer-aided diagnosis is a research area of increasing interest in third-level pediatric hospital care. The effectiveness of surgical treatments improves with accurate and timely information, and machine learning techniques have been employed to assist practitioners in making decisions. In this context, the prediction of the discharge diagnosis of new incoming patients could make a difference for successful treatments and optimal resource use. In this paper, a computer-aided diagnosis system is proposed to provide statistical information on the discharge diagnosis of a new incoming patient, based on the historical records from previously treated patients. The proposed system was trained and tested using a dataset of 1196 records, the dataset was coded according to the International Classification of Diseases, version 10 (ICD10). Among the processing steps, relevant features for classification were selected using the sequential forward selection wrapper, and outliers were removed using the density-based spatial clustering of applications with noise. Ensembles of decision trees were trained with different strategies, and the highest classification accuracy was obtained with the extreme Gradient boosting algorithm. A 10-fold cross-validation strategy was employed for system evaluation, and performance comparison was performed in terms of accuracy and F-measure. Experimental results showed an average accuracy of 84.62%, and the resulting decision tree learned from the experience in samples allowed it to visualize suitable treatments related to the historical record of patients. According to computer simulations, the proposed classification approach using XGBoost provided higher classification performance than other ensemble approaches, the resulting decision tree can be employed to inform possible paths and risks according to previous experience learned by the system. Finally, the adaptive system may learn from new cases to increase decisions’ accuracy through incremental learning.

An increasingly large community—mainly from emerging economies—collects and buys “click toys” (i.e., scalable and adjustable toys composed by many pieces); this type of toy has been used for educational purposes; low-cost toys can be obtained from second-hand cargoes of separate parts, but it requires solving an order-picking problem where the primary guideline is that pieces could fit together. This is a common concern of logistical applications with limited resources. In general, providers of this type of service implement a handmade routing system in which a route is selected based on the pieces and customization required. Using this system, it is difficult to generate effective routes to pick-up pieces; for example, in many situations, the same routes are used for a long time, and new provisions on traffic behavior or new routes available to ensure the correct supply of each piece are not considered. In this chapter, the authors propose an amalgam model using case-based reasoning and an algorithm based on the behavior of birds to solve the problem.

In this paper, we present an Android App designed to detect damage in tobacco leaves caused by the fungus of blue mold. This mobile application uses a classifier model which was built using a pattern recognition technique known as Artificial Neural Network. For the training and testing stages, a total of 40 images of tobacco leaves were used. The experimentation carried out shows that the developed model has an accuracy higher than 97% and there is no significant difference with a visual analysis carried out by experts in tobacco crop.

The classification of fresh fruits according to their ripeness is typically a subjective and tedious task; consequently, there is growing interest in the use of non-contact techniques such as those based on computer vision and machine learning. In this paper, we propose the use of non-intrusive techniques for the classification of Cape gooseberry fruits. The proposal is based on the use of machine learning techniques combined with different color spaces. Given the success of techniques such as artificial neural networks,support vector machines, decision trees, and K-nearest neighbors in addressing classification problems, we decided to use these approaches in this research work. A sample of 926 Cape gooseberry fruits was obtained, and fruits were classified manually according to their level of ripeness into seven different classes. Images of each fruit were acquired in the RGB format through a system developed for this purpose. These images were preprocessed, filtered and segmented until the fruits were identified. For each piece of fruit, the median color parameter values in the RGB space were obtained, and these results were subsequently transformed into the HSV and L*a*b* color spaces. The values of each piece of fruit in the three color spaces and their corresponding degrees of ripeness were arranged for use in the creation, testing, and comparison of the developed classification models. The classification of gooseberry fruits by ripening level was found to be sensitive to both the color space used and the classification technique, e.g., the models based on decision trees are the most accurate, and the models based on the L*a*b* color space obtain the best mean accuracy. However, the model that best classifies the cape gooseberry fruits based on ripeness level is that resulting from the combination of the SVM technique and the RGB color space.

In this work, the use of expert systems and hyperspectral imaging in the determination of coffee rust infection was evaluated. Three classifiers were trained using spectral profiles from different stages of infection, and the classifier based on a support vector machine provided the best performance. When this classifier was compared to visual analysis, statistically significant differences were observed, and the highest sensitivity of the selected classifier was found at early stages of infection.

The objective of this research was to develop a methodology for optimizing multilayer-per - ceptron-type neural networks by evaluating the effects of three neural architecture parame- ters, namely, number of hidden layers (HL), neurons per hidden layer (NHL), and activation function type (AF), on the sum of squares error (SSE). The data for the study were obtained from quality parameters (physicochemical and microbiological) of milk samples. Architec- tures or combinations were organized in groups (G1, G2, and G3) generated upon inter- spersing one, two, and three layers. Within each group, the networks had three neurons in the input layer, six neurons in the output layer, three to twenty-seven NHL, and three AF (tan-sig, log-sig, and linear) types. The number of architectures was determined using three factorial-type experimental designs, which reached 63, 2 187, and 50 049 combinations for G1, G2 and G3, respectively. Using MATLAB 2015a, a logical sequence was designed and implemented for constructing, training, and evaluating multilayer-perceptro n-type neural networks using parallel computing techniques. The results show that HL and NHL have a statistically relevant effect on SSE, and from two hidden layers, AF also has a significant effect; thus, both AF and NHL can be evaluated to determine the optimal combination per group. Moreover, in the three study groups, it is observed that there is an inverse relation- ship between the number of processors and the total optimization time.