Introducción: La neuropsicología del deporte ha tomado interés en el desarrollo procesos cognitivos de orden superior llamados funciones ejecutivas, estos se encuentran implicados en el desarrollo de tareas complejas, capacidad de adaptación y salud mental; evidencia reciente sugiere que estos procesos pueden ser estimulados a través de la actividad física. Sin embargo, el análisis de la relación entre estas variables puede verse sesgado con la utilización de métodos estadísticos clásicos debido a distintos supuestos teóricos y a las condiciones naturales de la investigación, por ello, la utilización de otras herramientas de análisis podría ser funcional para el estudio de estos procesos. Objetivo: Analizar la relación entre la actividad física y las funciones ejecutivas utilizando técnicas de inteligencia computacional. Metodología: Estudio cuantitativo, no experimental y transversal con una muestra de 140 estudiantes universitarios divididos en grupos de actividad física baja (n=47), media (n=40) y alta (n=53). Se utilizó el Inventario Internacional de Actividad Física (IPAQ) y la Batería Neuropsicológica de Funcionamiento Ejecutivo (BANFE-II). Resultados: Las técnicas de inteligencia computacional clasificaron el mejor rendimiento cognitivo con niveles más altos de actividad física en dos condiciones: 1) Perfil de desempeño cognitivo; 2) Puntuación total. Discusión: Los hallazgos corroboran la relación teórica establecida en otras investigaciones sobre cognición y actividad física, sin embargo, difieren en resultados con las técnicas de estadística clásica. Conclusiones: Los niveles de actividad física pueden determinar el desempeño cognitivo. Se muestra evidencias de la utilidad que la inteligencia computacional puede tener en la neuropsicología del deporte.

Computer vision and image processing have become relevant in recent years due to their capabilities to support different tasks in several areas. Image classification, segmentation, and estimation are relevant issues addressed using various techniques. Imaging estimation is very important and helpful in biological applications. This work proposes a new approach for estimating the damages in the livers of the Wistar rats, using high-resolution RGB images. Instead of using invasive methods to determine the level of damage, the proposal allows us to measure the damage in the livers. The proposal is based on Genetic Programming (GP), the paradigm of evolutionary computing, which has become relevant in recent years for image-processing tasks. It provides flexibility, which allows the use of image processing functions to extract meaningful information from raw images. Furthermore, it allows the configuration of the regression model by performing a hyperparameter tuning to improve estimation performance. The approach includes a new set of functions through which the regression model is configured. Additionally, a set of functions is included to change the color spaces of the images to extract meaningful features from them. The results demonstrate the effectiveness of our approach when making the hyperparameter tuning and the efficiency in dealing with different color spaces, thus achieving the promised results when estimating according to the R2, Mean Average Error (MAE), Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) indicators. The proposed method achieves values higher than 0.5 of R2 and lower than 0.51 of MSE, using different regression models. Additionally, the approach demonstrates that image preprocessing is necessary for improving the model’s performance, which is better than only using raw data where the values of RMSE are greater than 1.5. The lowest MSE value of our proposed method was 0.51, outperforming the methods without preprocessing.

The common bean is a widely cultivated food source. Many domesticated species of common bean varieties, known as landraces, are cultivated in Mexico by local farmers, exhibiting various colorations and seed mixtures as part of agricultural practices. In this work, we propose a methodology for classifying bean landrace samples using three two-dimensional histograms with data in the CIE L*a*b* color space while additionally integrating chroma (C*) and hue (h°) to develop a new proposal from histograms, employing deep learning for the classification task. The results indicate that utilizing three histograms based on L*, C*, and h° brings an average accuracy of 85.74 ± 2.37 compared to three histograms using L*, a*, and b*, which reported an average accuracy of 82.22 ± 2.84. In conclusion, the new color characterization approach presents a viable solution for classifying common bean landraces of both homogeneous and heterogeneous colors.

Recently, Convolutional Neural Networks (CNNs) have performed well in pattern recognition tasks such as image segmentation. However, designing a CNN for a particular dataset requires experience in choosing from feature extraction and aggregation operators. Neural Architecture Search (NAS) is automating the design of neural networks for a specific task. By leveraging its syntax tree representation, genetic programming (GP) can be adapted as a search strategy in the NAS (ENAS) problem. Solutions evaluation implies the models' weight optimization, making ENAS computationally expensive. Nonetheless, surrogate models can reduce the costly evaluations by predicting fitness solutions. Generating a surrogate model from variable-size tree-based GP solutions usually requires a tree-to-structured data conversion to use standard baseline regression models. This paper compares several surrogate models used to assist GP-based algorithms. Also, phenotype-level features are proposed to be included in converting syntax-tree solutions to sequence-vectors. The surrogate models are applied to optimize U-Net-type networks, showing promising results in reducing the number of costly evaluations and the execution time without affecting the segmentation performance.

Convolutional Decision Trees (CDTs) are machine learning models utilized as interpretable methods for image segmentation. Their graphical structure enables a relatively simple interpretation of how the tree successively divides the image pixels into two classes, distinguishing between objects of interest and the image background. Several techniques have been proposed to induce CDTs. However, they have primarily been focused on analyzing grayscale images due to the computational cost of the Differential Evolution (DE) algorithm, which is employed in these techniques. This paper proposes a generalization of the induction process of a CDT with the DE algorithm using color images, implementing two techniques to reduce the computational time and memory employed in the induction process: the median selection technique and a memory of previously evaluated solutions. The first technique is applied to select a representative sample of pixels from an image for the model’s training process, and the second technique is implemented to reduce the number of evaluations in the fitness function considered in the DE process. The efficacy of these techniques was evaluated using the Weizmann Horse and DRIVE datasets, resulting in favorable outcomes in terms of the segmentation performance of the induced CDTs, and the processing time and memory required for the induction process.

This study explores a novel encoding approach for optimizing Liquid State Machines (LSMs) in a Neuroevolution (NE) context. By leveraging a Genetic Algorithm (GA), two components of LSM design, neuron configurations and neuron positions, are used. Three variants were evaluated to assess their individual contributions: the complete proposal considering both components, one encoding with only neuron configurations, and another focusing on neuron positions. Experiments were conducted on varying complexity synthetic classification tasks, demonstrating that neuron configurations significantly influence performance, while neuron positions alone were less effective. Statistical analysis using the Shapiro-Wilk, Kruskal-Wallis, and Dunn's post hoc tests validated these findings. The results highlight the importance of configuration-driven encodings in LSM optimization and suggest the need for refined evolutionary strategies to exploit positional information better.

Image segmentation is a challenging task due to image variations, such as illumination, background, noise, and others. There are several segmentation methods, but the requirement of prior knowledge and parameter setting makes it hard to perform a good segmentation, especially in medical images where an expert is needed to make the segmentation accordingly with the prior knowledge to determine the area of interest. In this work, we are focused on the feature construction to make an image segmentation of computerized tomography scans of lungs affected by COVID-19. Genetic Programming (GP) is used to evolve a program to extract and construct features from the image to make a segmentation where the target is to find the affected area. The flexibility that offers GP allows us to face the segmentation task and know which functions are used in the final program, leading to an interpretable solution. The results of the experiments demonstrate that GP is capable of extracting and constructing features from the Computerized Tomography images to perform the segmentation of lungs affected by COVID-19, achieving values of 0.59 of F-1 - score metric to measure the segmentation performance. Furthermore, the experimental results determine the appropriate parameters for the evolutionary process.

Neural Architecture search (NAS) aims to automate the design process of Deep Neural Networks, reducing the Deep Learning (DL) expertise required and avoiding a trial-and-error process. Nonetheless, one of the main drawbacks of NAS is the high consumption of computational resources. Consequently, efficient evaluation methods (EEMs) to assess the quality of candidate architectures are an open research problem. This work tests various EEMs in the Deep Genetic Algorithm (DeepGA), including early stopping, population memory, and training-free proxies. The Fashion MNIST, CIFAR-10, and CIFAR-100 datasets were used for experimentation. The results show that population memory has a valuable impact on avoiding repeated evaluations. Additionally, early stopping achieved competitive performance while significantly reducing the computational cost of the search process. The training-free configurations using the Logsynflow and Linear Regions proxies, as well as a combination of both, were only partially competitive but dramatically reduced the search time. Finally, a comparison of the architectures and hyperparameters obtained with the different algorithm configurations is presented. The training-free search processes resulted in deeper architectures with more fully connected layers and skip connections than the ones obtained with accuracy-guided search configurations.

The classification of bean landraces based on their coloration is of particular interest, as the color of these plants is associated with the nutritional components present in their seeds. In this paper, the authors propose a procedure to identify the colors of heterogeneous color bean landraces based on the information from their digital images. The proposed methodology employs a three-dimensional histogram representation of the estimated color, expressed in the CIE L*a*b* color space, with an unsupervised learning method called the Gaussian Mixture Model. This approach facilitates the acquisition of representative information for the colors of a bean landrace, represented as points in the CIE L*a*b* color space. Furthermore, the K-nn method can be trained with these punctual representations to identify colors, yielding satisfactory results on landraces with homogeneous and heterogeneous seeds.

Within the medical field, computer vision has an important role in different tasks, such as health anomaly detection, diagnosis, treatment, and monitoring medical conditions. Image segmentation is one of the most used techniques for medical support to identify regions of interest in different organs. However, performing accurate segmentation is difficult due to image variations. In this way, this work proposes an automated multiple-feature construction approach for image segmentation, working with magnetic resonance images, computed tomography, and RGB digital images. Genetic programming is used to automatically create and construct pipelines to extract meaningful features for segmentation tasks. Additionally, a co-evolution strategy is proposed within the evolution process to increase diversity without affecting segmentation performance. The segmentation is addressed as a pixel classification task; in this way, a wrapper approach is used, and the classification model’s segmentation performance determines the fitness. To validate the effectiveness of the proposed method, four datasets were used to measure the capability of the proposal to deal with different types of medical images. The results demonstrate that the proposal achieves values of the DICE similarity coefficient of more than 0.6 in MRI and C.T. images. Additionally, the proposal is compared with SOTA GP-based methods and the convolutional neural networks used within the medical field. The method proposed outperforms these methods, achieving improvements greater than 20% in DICE, specificity, and sensitivity. Additionally, the qualitative results demonstrate that the proposal accurately identifies the region of interest.

Background/Objectives: Tuberculosis (TB), caused by Mycobacterium tuberculosis (M. tuberculosis), remains a leading cause of death from infectious diseases globally. The treatment of active TB relies on first- and second-line drugs, however, the emergence of drug resistance poses a significant challenge to global TB control efforts. Recent advances in whole-genome sequencing combined with machine learning have shown promise in predicting drug resistance. This study aimed to evaluate the performance of four machine learning models in classifying resistance to ethambutol, isoniazid, and rifampicin in M. tuberculosis isolates. Methods: Four machine learning models—Extreme Gradient Boosting Classifier (XGBC), Logistic Gradient Boosting Classifier (LGBC), Gradient Boosting Classifier (GBC), and an Artificial Neural Network (ANN)—were trained using a Variant Call Format (VCF) dataset preprocessed by the CRyPTIC consortium. Three datasets were used: the original dataset, a principal component analysis (PCA)-reduced dataset, and a dataset prioritizing significant mutations identified by the XGBC model. The models were trained and tested across these datasets, and their performance was compared using sensitivity, specificity, Precision, F1-scores and Accuracy. Results: All models were applied to the PCA-reduced dataset, while the XGBC model was also evaluated using the mutation-prioritized dataset. The XGBC model trained on the original dataset outperformed the others, achieving sensitivity values of 0.97, 0.90, and 0.94; specificity values of 0.97, 0.99, and 0.96; and F1-scores of 0.93, 0.94, and 0.92 for ethambutol, isoniazid, and rifampicin, respectively. These results demonstrate the superior accuracy of the XGBC model in classifying drug resistance. Conclusions: The study highlights the effectiveness of using a binary representation of mutations to train the XGBC model for predicting resistance and susceptibility to key TB drugs. The XGBC model trained on the original dataset demonstrated the highest performance among the evaluated models, suggesting its potential for clinical application in combating drug-resistant tuberculosis. Further research is needed to validate and expand these findings for broader implementation in TB diagnostics.

A pigment of great interest is the anthocyanins. It is due to the nutritional benefits discovered in various foods, such as common beans. In this work, we report the estimation of anthocyanins in homogeneous colored bean landraces using neuroevolution. Two neuroevolution techniques, NEAT and DeepGA, were implemented to find this task’s suitable neural network structure. Both techniques were compared against a Convolutional Neural Network (CNN) experimentally developed called AnthEst-Net architecture, which found competitive results in anthocyanin estimation. The input data of the network architectures were two-color characterizations, two-dimensional histograms, and data vectors. The accuracies obtained on the test set in HSI color space were 85.38 ± 11.77 and 87.89 ± 9.67 for DeepGA and AnthEstNet architecture, respectively. Regarding CIE L*a*b* color space, DeepGA obtained an accuracy of 86.85 ± 11.08, while AnthEstNet got 87.08 ± 14.19. Results suggest that the architecture reported by DeepGA is suitable for anthocyanins estimation.

Behavioral neuropharmacology is an area of neuroscience, which is responsible for the study of behavioral modifications through the administration of substances, treatments or experimental manipulations. Particularly to determine the effect on behavior, this area uses classic statistical techniques for comparing measures of central tendency; the analysis of variables is mostly carried out in a univariate manner, where the interpretation of the results obtained is often limited. There are other areas that also provide tools for data analysis, such as computational learning, through prediction models we can determine the characteristic behavioral patterns of each treatment administered. In the present study, computational learning data analysis techniques were used, specifically, supervised machine learning applied to a behavioral neuropharmacology experiment, where 3 doses of allopregnanolone (0.5, 1, and 2 mg) were evaluated in maze tests. Raised arms and motor activity test. We identified with classical statistical methods that the 2 mg dose of allopregananolone has an anxiolytic-type effect, similar to that exerted by the reference drug diazepam. Additionally, with computational learning methods, we can identify the characteristic patterns of each treatment based on the combination of the variables of both behavioral tests, likewise, we demonstrate with mathematical support the most important variables for the identification of anxioselective effects. In conclusion, computational learning methods promote enrichment in the results of neuropharmacology reflected in the characteristic patterns that are modified by the administration of different drugs, and provide foundations to support the importance of the most relevant variables of behavioral tests.

Wrapper approaches for feature selection are known for their high performance, but the drawback of high computational cost is presented. This work proposes using cost-reduction mechanisms applied to the permutational-based Differential Evolution (DE-FS PM ) algorithm for feature selection. Two proposals considering fixed and incremental sampling fraction strategies are considered to reduce the cost of evaluating an individual. A memory mechanism for avoiding repeated evaluations is included. The success-history parameter adaptation for Differential Evolution (SHADE) procedure adapted to the permutational search space is applied in two additional proposals. Eighteen datasets were used for experimentation. The fixed sampling fraction proposal with the memory mechanism reached competitive accuracy results while requiring less computational time. The sampling strategies could effectively reduce the number of dataset instances used for evaluation. In addition, the memory mechanism avoids a fraction of the evaluations in the search process. The results show that two simple mechanisms can effectively decrease the computational cost of a wrapper approach for feature selection without diminishing its performance.

Breast cancer (BC) has become a global health problem, ranking first in incidence and fifth in mortality in women around the world. Although there are some diagnostic methods for the disease, these are not sufficiently effective and are invasive. In this work, we discriminated between patients without breast pathology (BP), with benign BP, and with BC based on the band patterns obtained from Western blot strip images of the autoantibody response to antigens of the T47D tumor line using and comparing supervised machine learning techniques to have a sensitive and accurate method. When comparing the aforementioned machine learning techniques, it was found that by obtaining a convolutional neural network architecture from a neuroevolution algorithm, it is possible to automatically discriminate with a classification accuracy of 90.67% between patients with cancer and with/without BP. In the case of discrimination between patients with cancer and without BP, a classification accuracy of 96.67% was obtained with the K-NN algorithm and 95.13% with the convolutional neural network obtained using a neuroevolution algorithm, although these results are not statistically significant. It is concluded that the convolutional neural network obtained by neuroevolution is the method with the best performance with respect to those evaluated in this work.

Behavioral neuropharmacology, a branch of neuroscience, uses behavioral analysis to demonstrate treatment effects on animal models, which is fundamental for pre-clinical evaluation. Typically, this determination is univariate, neglecting the relevant associations for understanding treatment effects in animals and humans. This study implements regression trees and Bayesian networks from a multivariate perspective by using variables obtained from behavioral tests to predict the time spent in the open arms of the elevated arm maze, a key variable to assess anxiety. Three doses of allopregnanolone were analyzed and compared to a vehicle group and a diazepam-positive control. Regression trees identified cut-off points between the anxiolytic and anxiogenic effects, with the anxiety index standing out as a robust predictor, combined with the percentage of open-arm entries and the number of entries. Bayesian networks facilitated the visualization and understanding of the interactions between multiple behavioral and biological variables, demonstrating that treatment with allopregnanolone (2 mg) emulates the effects of diazepam, validating the multivariate approach. The results highlight the relevance of integrating advanced methods, such as Bayesian networks, into preclinical research to enrich the interpretation of complex behavioral data in animal models, which can hardly be observed with univariate statistics.

Time series classification is a supervised task in the field of temporal data mining. Time series naturally tend to be highly dimensional, requiring the use of reduction techniques such as discretization. eMODiTS is a data-driven method for symbolically discretizing time series, which determines the best scheme by modifying the number of time (word segments) and values (alphabet) cuts, generating a unique alphabet set for every word segment. However, due to the high computational cost required, a surrogate model is incorporated to minimize this cost, using the K-Nearest Neighbors approach for regression and Dynamic Time Warping (DTW) as the similarity measure. Results suggest that the surrogate model effectively estimates the objective functions’ values similarly to the original ones, leading to similar classification rates. It is validated with the statistical test where there is no significant statistical difference between the surrogate and original models. The surrogate model produces modified acceptance index ( $$d_j$$ ) values regarding predicting ability, indicating that the predictive performance is on average. On the other hand, the Mean Squared Error (MSE) consistently stays below 0.15, demonstrating that even when surrogate models cannot estimate the same values as the original model, the similarity of the values remains clear.

Drought prediction is a complex phenomenon that impacts human activities and the environment. For this reason, predicting its behavior is crucial to mitigating such effects. Deep learning techniques are emerging as a powerful tool for this task. The main goal of this work is to review the state-of-the-art for characterizing the deep learning techniques used in the drought prediction task. The results suggest that the most widely used climate indexes were the Standardized Precipitation Index (SPI) and the Standardized Precipitation Evapotranspiration Index (SPEI). Regarding the multispectral index, the Normalized Difference Vegetation Index (NDVI) is the indicator most utilized. On the other hand, countries with a higher production of scientific knowledge in this area are located in Asia and Oceania; meanwhile, America and Africa are the regions with few publications. Concerning deep learning methods, the Long-Short Term Memory network (LSTM) is the algorithm most implemented for this task, either implemented canonically or together with other deep learning techniques (hybrid methods). In conclusion, this review reveals a need for more scientific knowledge about drought prediction using multispectral indices and deep learning techniques in America and Africa; therefore, it is an opportunity to characterize the phenomenon in developing countries.