SSSalvador Santos-Romero

Salvador Santos-Romero

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Salvador Santos-Romero has a total of 8 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    A novel data-driven statistical learning algorithm for neural networks applied to language translation

    by Juan German Hernández Sánchez, José Manuel Cruz Olguín, Salvador Santos Romero, Anilú Franco Árcega, Virgilio López Morales, Manuel Alejandro Ojeda Misses

    This article presents a novel statistical data-driven learning algorithm for neural network training, designed for low-resource data scenarios. The proposed method employs statistical descriptors, including the sample mean, standard deviation, z-score, and normal distribution, to estimate the probability of prediction error and guide the adaptive updating of weights and biases. By incorporating error normalization and probabilistic mappings, the algorithm provides a statistically guided learning mechanism that accounts for signal variability while promoting stable convergence. Contrasting conventional approaches that rely primarily on gradient information, the proposed framework uses probabilistic information derived from the distribution of error to regulate parameter adaptation. The methodology is validated using Nahuatl–Spanish lexical translation task implemented with neural networks. • Proposes a statistical–probabilistic learning algorithm that uses error distributions instead of relying only on gradients.• The method is intended for low-resource scenarios and shows stable convergence under the evaluated conditions.• Validated on a Nahuatl–Spanish lexical translation task using a multilayer neural network with comparable performance than classical methods.

  • Open Access

    Algoritmo estadístico para el aprendizaje en redes neuronales para la traducción hñähñu-español

    by José Manuel Cruz-Olguín, Salvador Santos-Romero, Virgilio López-Morales, Manuel Alejandro Ojeda-Misses

    La traducción automática de lenguas indígenas de bajo recurso enfrenta importantes desafíos derivados de la escasez de corpus digitales, la variación dialectal y la complejidad lingüística. El objetivo de este trabajo fue desarrollar y evaluar un traductor automático hñähñu–español basado en una red neuronal perceptrón multicapa entrenada mediante un algoritmo estadístico–probabilístico (EyP), fundamentado en el Teorema del Límite Central. Se generó una metodología que permitiera realizar el ajuste de pesos y sesgos a partir de probabilidades de error estimadas estadísticamente, sin emplear derivadas explícitas, diferenciándolo de los algoritmos clásicos de entrenamiento neuronal como Backpropagation y Levenberg–Marquardt. Se utilizó una estrategia de vectorización a nivel de caracter para representar palabras del hñähñu como vectores numéricos de longitud fija, lo que permitió capturar regularidades fonético–estructurales entre ambas lenguas. El desempeño del traductor se evaluó utilizando un corpus léxico de 145 palabras, métricas estadísticas objetivas a través del error cuadrático medio (ECM), validación cruzada k-fold y comparaciones directas con algoritmos de entrenamiento tradicionales. Los resultados mostraron que el algoritmo EyP alcanzó menores ECM y una convergencia más estable, especialmente en escenarios de datos limitados, aunque con un mayor número de épocas. En conjunto, los hallazgos confirman que el enfoque EyP propuesto constituye una alternativa robusta y viable para la traducción automática hñähñu–español, contribuyendo a la preservación de lenguas indígenas.

  • Open Access

    Dynamic Evasion Point System for Mobile Robots

    by Salvador Santos-Romero, Virgilio López_Morales, Julio C. Ramos-Fernández, Manuel A. Ojeda-Misses

    In this paper, a novel local path planner based on a complete Obstacle Avoidance Algorithm (OAA) and a Proportional-Derivative (PD) controller is developed for Mobile Robots (MR). The main contribution of this work is a local planner that generates global evasion points instead of just angles, and uses a normalized cost function that weights between the total Euclidean route distance and the orientation turning cost, separating the evasion algorithm from the pose tracking controller. ”Individual obstacles detected by the LiDAR (Light Detection and Ranging sensor) are classified by measuring the distance between their corners. The OAA determines whether the mobile robot can drive through them or considers them as a single obstacle. A target point (an evasion point or the final goal) is selected by minimizing a normalized cost function. Various real experiments were carried out comparing the VPH and our method to illustrate the advantages and efficiency in both scenarios, achieving successful evasion in complex environments, such as ’U’ structures, and maintaining stable CPU usage. Our method uses a PD controller to drive both wheels and allows the mobile robot to reach its destination point in a smooth and efficient manner.