ATArturo Téllez-Velázquez

Arturo Téllez-Velázquez

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Arturo Téllez-Velázquez has a total of 24 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Semantic Reasoning Using Standard Attention-Based Models: An Application to Chronic Disease Literature

    by Yalbi Itzel Balderas-Martínez, José Armando Sánchez-Rojas, Arturo Téllez-Velázquez, Flavio Juárez Martínez, Raúl Cruz-Barbosa, Enrique Guzmán-Ramírez, Iván García-Pacheco, Ignacio Arroyo-Fernández

    Large-language-model (LLM) APIs demonstrate impressive reasoning capabilities, but their size, cost, and closed weights limit the deployment of knowledge-aware AI within biomedical research groups. At the other extreme, standard attention-based neural language models (SANLMs)—including encoder–decoder architectures such as Transformers, Gated Recurrent Units (GRUs), and Long Short-Term Memory (LSTM) networks—are computationally inexpensive. However, their capacity for semantic reasoning in noisy, open-vocabulary knowledge bases (KBs) remains unquantified. Therefore, we investigate whether compact SANLMs can (i) reason over hybrid OpenIE-derived KBs that integrate commonsense, general-purpose, and non-communicable-disease (NCD) literature; (ii) operate effectively on commodity GPUs; and (iii) exhibit semantic coherence as assessed through manual linguistic inspection. To this end, we constructed four training KBs by integrating ConceptNet (600k triples), a 39k-triple general-purpose OpenIE set, and an 18.6k-triple OpenNCDKB extracted from 1200 PubMed abstracts. Encoder–decoder GRU, LSTM, and Transformer models (1–2 blocks) were trained to predict the object phrase given the subject + predicate. Beyond token-level cross-entropy, we introduced the Meaning-based Selectional-Preference Test (MSPT): for each withheld triple, we masked the object, generated a candidate, and measured its surplus cosine similarity over a random baseline using word embeddings, with significance assessed via a one-sided t-test. Hyperparameter sensitivity (311 GRU/168 LSTM runs) was analyzed, and qualitative frame–role diagnostics completed the evaluation. Our results showed that all SANLMs learned effectively from the point of view of the cross entropy loss. In addition, our MSPT provided meaningful semantic insights: for the GRUs (256-dim, 2048-unit, 1-layer): mean similarity (μsts) of 0.641 to the ground truth vs. 0.542 to the random baseline (gap 12.1%; p<10−180). For the 1-block Transformer: μsts=0.551 vs. 0.511 (gap 4%; p<10−25). While Transformers minimized loss and accuracy variance, GRUs captured finer selectional preferences. Both architectures trained within <24 GB GPU VRAM and produced linguistically acceptable, albeit over-generalized, biomedical assertions. Due to their observed performance, LSTM results were designated as baseline models for comparison. Therefore, properly tuned SANLMs can achieve statistically robust semantic reasoning over noisy, domain-specific KBs without reliance on massive LLMs. Their interpretability, minimal hardware footprint, and open weights promote equitable AI research, opening new avenues for automated NCD knowledge synthesis, surveillance, and decision support.


  • Relaxed Differential Evolution Algorithm

    by Prometeo Cortés-Antonio, Arturo Téllez-Velázquez, Raúl Cruz-Barbosa, Oscar Castillo


  • Comparative Analysis of Clustering Methods for Fuzzy Classifiers Simplification

    by Luis Diego Baños-Zamora, Arturo Téllez-Velázquez, Rosebet Miranda-Luna

    One problem of fuzzy systems for classification tasks is the exponential growth of rule generation, which translates into excessively long processing times. Different proposals in the literature address this problem through exhaustive rule reduction techniques, which achieve competitive results compared to conventional classifiers by reducing their computational complexity as well. This paper proposes a methodology that is comprised of two stages: 1) a clustering technique that helps identify the structure of an initial low-accurate classifier and 2) a differential evolution parameter identification stage that takes that low-accurate classifier and refines it to obtain a high-accurate classifier. In this way, a comparative analysis among several clustering methods is performed, allowing the users to select a reduced rule set and avoid the use of traditional rule search algorithms. The results show that the Gaussian Mixture Model is the most suitable clustering technique to identify the structure of fuzzy classifiers, since it provides the corresponding fuzzy model with highly competitive classification performance compared to other state-of-the-art methods.

  • Open Access

    Optimal Position Fuzzy Control of an Underactuated Robotic Finger

    by Francisco J. Espinosa Garcia, Esther Lugo-González, Arturo Telléz-Velázquez, Manuel Arias-Montiel, Marco Ceccarelli

    In this paper, the optimal position control of an underactuated robotic finger is presented. Two trajectories, one for the proximal and the other for the medial phalanx, are proposed in order to emulate the fingers flexion/extension movements. A Mandani fuzzy control is proposed due to the lack of a precise dynamical model of the system. In order to obtain the control parameters, an optimization strategy based on the membership functions is applied. Genetic algorithms (GA) are commonly used as an optimization method in diverse applications; however, in this case, the use of an autoadaptive differential evolution method is proposed in order to obtain a superior convergence behavior. Simulations of the virtual prototype are carried out using MATLAB/Simulink software to display the trajectory tracking. The results show that the maximum error between the proposed and obtained trajectories is 3.1352E04rad.


  • Open Access

    Patrones que identifican a estudiantes universitarios desertores aplicando minería de datos educativa

    by Argelia Berenice Urbina-Nájera, Arturo Téllez-Velázquez, Raúl Cruz Barbosa

    En este trabajo se presenta un análisis de las características más relevantes de un potencial desertor universitario, mediante la aplicación de algoritmos de minería de datos educativa. Se utilizó un conjunto de datos de 10 635 instancias, adquiridas en el período 2014-2019, de 53 programas de licenciatura de una institución privada del estado de Puebla (México). Los resultados muestran que el modelo obtenido por los árboles de decisión ofrece mayor desempeño que otros algoritmos, así como una fácil interpretación de éste mediante reglas de decisión. Además, el rendimiento del modelo es mejor que otros modelos relacionados en la literatura aplicados al mismo problema. Los métodos de selección de características permitieron encontrar los atributos más importantes que identifican a un potencial desertor, tales como: el período, el último semestre cursado, créditos cursados, asistencia, materias reprobadas y programa. Utilizando los atributos y reglas de decisión encontradas se podrían crear mecanismos que favorezcan la prevención de la deserción.


  • Open Access

    On the Feasibility of Fast Fourier Transform Separability Property for Distributed Image Processing

    by Arturo Téllez-Velázquez, Raúl Cruz-Barbosa

    Given the high algorithmic complexity of applied-to-images Fast Fourier Transforms (FFT), computational-resource-usage efficiency has been a challenge in several engineering fields. Accelerator devices such as Graphics Processing Units are very attractive solutions that greatly improve processing times. However, when the number of images to be processed is large, having a limited amount of memory is a serious problem. This can be faced by using more accelerators or using higher-capability accelerators, which implies higher costs. The separability property is a resource in hardware approaches that is frequently used to divide the two-dimensional FFT work into several one-dimensional FFTs, which can be simultaneously processed by several computing units. Then, a feasible alternative to address this problem is distributed computing through an Apache Spark cluster. However, determining the separability-property feasibility in distributed systems, when migrating from hardware implementations, is not evident. For this reason, in this paper a comparative study is presented between distributed versions of two-dimensional FFTs using the separability property to determine the suitable way to process large image sets using both Spark RRDs and DataFrame APIs.


  • A Spark image processing toolkit

    by Arturo Téllez‐Velázquez, Raúl Cruz‐Barbosa

    Summary: The main drawback of conventional tools for digital image processing is the long processing time due to the high complexity of their algorithms. This gets worse when these algorithms need to be sequentially processed with large image sets. To alleviate part of this situation, this paper introduces a general‐purpose tool for massively processing large digital image sets by using Apache Spark. The proposed tool allows users to extract the image rasters and store them in any of Spark's basic distributed data representations, namely, Resilient Distributed Datasets (RDD) and DataFrame (DF), to treat all the subsequent image operations as RDD/DF transformations. Our experiments reveal that, with our proposal, it is possible to schedule and execute distributed image processing tasks in less time, in comparison with another Spark‐based massive image processing tool. In these experiments, we applied several algorithms to 25 000 images (the MIRFLICKR‐25000 set), reaching a maximum speedup of 54x. In addition, it was discovered that the number of images also influences the speedup, as the cluster memory is fully occupied. Therefore, we can claim that, using our proposal, more complex image processing workflows can be built and applied massively to large image sets, achieving competitive speedups.


  • Open Access

    Análisis exploratorio para la caracterización de la adicción a la cocaína a través del aprendizaje computacional

    by Arturo Téllez-Velázquez, Eduardo A. Garza-Villareal, Jorge J. González-Olvera, Raúl Cruz-Barbosa

  • A CUDA‐streams inference machine for non‐singleton fuzzy systems

    by Arturo Téllez‐Velázquez, Raúl Cruz‐Barbosa

    Summary: The inferences computed on a non‐singleton fuzzy system represent a processing challenge when the number of linguistic variables and terms in a rule set are large, since they involve the use of a lot of sequential vector operations. To alleviate part of the inference process, a general‐purpose CUDA fuzzy tool for building non‐singleton fuzzy systems is presented in this paper. Here, we introduce an inference machine architecture that processes string‐based rules and concurrently executes them, based on an execution plan created by a fuzzy rule scheduler. This operation breaks down then−ary fuzzy operations in several binary fuzzy operations which can be executed in several streams and stages. As a result, this approach lets a system both take advantage of the parallel nature of rule sets and achieve competitive speed‐up ratios without losing generality, when selecting large numbers of linguistic variables, linguistic terms, and rules. The object‐oriented nature of the proposed tool makes it an easy way to build fuzzy systems without having a deep knowledge of its architecture, as it is shown in the results of two case studies for testing the fuzzy system operation limit and detecting edges in digital images.


  • A Feasible Genetic Optimization Strategy for Parametric Interval Type-2 Fuzzy Logic Systems

    by Arturo Téllez-Velázquez, Herón Molina-Lozano, Luis A. Villa-Vargas, Raúl Cruz-Barbosa, Esther Lugo-González, Ildar Z. Batyrshin, Imre J. Rudas