Ciencias Exactas y Ciencias de la Salud
Permanent URI for this collectionhttps://hdl.handle.net/11285/551039
Pertenecen a esta colección Tesis y Trabajos de grado de las Maestrías correspondientes a las Escuelas de Ingeniería y Ciencias así como a Medicina y Ciencias de la Salud.
Browse
Search Results
- Hardware-aware neural architecture search for enhancing text generation(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2025-06) Sánchez Miranda, Israel; Sosa Hernández, Víctor Adrián; emipsanchez; Castillo Juárez, Esteban; Ortiz Bayliss, José Carlos; Juárez Gambino, Joel Omar; School of Engineering and Sciences; Campus Estado de México; Pescador Rojas, MiriamIn recent years, neural network optimization has become critical in Natural Language Processing (NLP) tasks. However, manual tuning processes are time-consuming and heavily influenced by the designer’s prior knowledge, limiting the exploration of alternative architecture designs. Consequently, only a narrow subset of neural network architectures is typically considered for tasks such as text generation. Furthermore, neural network tuning requires specialized expertise, posing a barrier for non-experts and hindering broader innovation in the field. This research addresses these challenges by implementing a specialized Hardware-Aware Neural Architecture Search (HW-NAS) methodology, tailored specifically for text generation tasks under resource-constrained environments. The proposed NAS approach leverages a compact, efficient search space encoding key transformer architectural components, while adopting multi-objective optimization to simultaneously maximize text generation quality, measured via the METEOR score, and minimize the parameter count to enhance hardware adaptability. Two different evolutionary-based NAS strategies were explored: a custom Lexicographic Evolutionary Strategy (LexSMS-MODES) and SMS-EMOA, focusing on balancing exploration, exploitation, and computational efficiency. Experimental evaluations were conducted in both unconstrained environments and constrained hardware platforms. The optimized architectures demonstrated consistent improvements over the baseline model across multiple performance measures, including BLEU, ROUGE, and GLEU. Notably, METEOR scores showed values close to 0.72 in unconstrained settings. Although significant performance degradation was observed under constrained environments (approximately 57%–59% reduction in METEOR scores), the discovered models maintained a competitive edge when compared to several state-ofthe-art light-weight and NAS-based solutions. Hardware-aware evaluations revealed that NAS-generated models achieved substantial reductions in memory usage, GPU load, and CPU frequency deltas, despite not explicitly optimizing hardware indicators during the search. Statistical tests confirmed the stability of the discovered models across multiple hardware performance metrics. Comparisons against external works showed that while the proposed method successfully produced light-weight and efficient architectures, there remains room for improvement regarding inference latency and hardware adaptation strategies.
- Optimization of kinetic and operating parameters in bioreactors using evolutionary algorithms(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-11) Barrera Hernández, Gonzalo Irving; Sosa Hernández, Víctor Adrián; emipsanchez; Alfaro Ponce, Mariel; Aranda Barradas, Juan Silvestre; Corrales Muñoz, David Camilo; School of Engineering and Sciences; Campus Estado de México; Gómez Acata, Rigel ValentínBioreactors play a role in creating biological products such as medicines and biofuels by care fully controlling factors such as substrate levels and temperature within them to obtain optimal production results, bioreactor production process poses a challenge that poses a challenge to engineers due to the intricate setup involved. In the field of microbiology and biotechnology, conventional approaches such as the Monod model, logistic growth models, and fed-batch techniques have been employed to predict and improve the growth conditions of microor ganisms and the production of proteins of interest in fermenters. However, these approaches could face challenges when they encounter nonlinear systems and conflicting objectives. To address these challenges, our suggestion is to approach the configuration of factors in bioreactors as an optimization problem using an evolutionary algorithm that can improve the effectiveness and quality of the operating process. The objective of this study is to in vestigate and create a pipeline that integrates evolutionary algorithms to solve multi-objective and scalar optimization problems, aimed at identifying kinetic and critical parameters within a bioreactor system. The optimization process involves, in the first stage, a least squares ap proach that considers product, biomass, dissolved oxygen, and substrate concentrations as objectives, with the kinetic parameters (e.g., maximum specific growth rate and substrate affinity) serving as decision variables. The second stage focuses solely on maximizing the amount of produced product, specifically biomass, using critical operational variables, such as feed rate and aeration, as decision variables. The research employs Escherichia coli as a microorganism that has been genetically al tered to produce orange fluorescent protein (OFP) to test the validity of improvement frame works. Initially, in the simulation and process tuning phase, experimental information, from batch cultures, is used to accurately determine the factors. Later, in the fed-batch phase, the application of an algorithm is used to optimize biomass yield while considering operational constraints such as oxygen levels and maximum reactor volume. The findings show that this method accurately calculates factors during the fed-batch phase and efficiently increases biomass production in the continuous fed phase. The use of algorithms such as multiple NSGA-III and single-objective genetic algorithms provides valuable benefits when dealing with intricate bioreactor configurations that have conflicting objectives such as managing substrate consumption and improving production yield. This approach has promising prospects for improving the accuracy and efficiency of bioprocess optimization, while increasing its scalability, in the field of biotechnology in the future.
- An improved multi-objective optimization problem model for enhancing UAV path planning(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024) Castañon Guerrero, Franco; Sosa Hernández, Víctor Adrián; mtyahinojosa; Yee Rendón, Arturo; Estrada Delgado, Mario Iván; Escuela de Ingeniería y Ciencias; Campus Estado de México; Becerril Gómez, Jorge AntonioUnmanned Aerial Vehicles (UAVs) have become crucial in various industries, such as agriculture, construction and mining, infrastructure inspection, environmental monitoring, and emergency response. These diverse applications underscore the importance of UAV drone path planning for enhancing efficiency and safety. This work builds upon the study presented by Wang et al., highlighting limitations in environmental modeling. The failure to accurately replicate the environmental conditions can be attributed to insufficient documentation of the modeling methodology, hindering the repeatability and robustness of the findings. Critiques also target the fitness functions lacking theoretical grounding. The Threat Index assesses flight smoothness but lacks clear operational descriptions, while the Concealment Index evaluates safety but suffers from unclear procedures. There is a need for an improved and accurate model. Our contribution introduces two new functions for UAV path planning, optimizing the two distinct aspects: the Threat and Concealment of the trajectory. The first proposed function focuses on the distance of the path to the surface, incorporating altitude variations and terrain features to minimize deviations from the surface. The second one addresses angular preferences, minimizing deviations from a straight-line trajectory to reduce the impact of inertia on UAV dynamics. The integration of the new objective functions contributes to a multi-objective optimization framework, balancing considerations of path proximity to the surface and path linearity for enhanced UAV path planning performance. Our framework involved conducting four test scenarios with distinct points of origin and goals, utilizing the SMS-EMOA algorithm to find the best path. Each experiment was characterized by unique initial and terminal coordinates, allowing for a comprehensive evaluation across diverse scenarios. The evolutionary algorithms were configured with specific parameters to balance computational efficiency with optimization robustness. Additionally, 30 independent runs were performed for each scenario, comparing the two sets of objective functions to capture the general behavior of each one. The success of the experiments was measured by the convergence of the algorithms towards Pareto-optimal solutions, demonstrating adaptability and effectiveness across varied spatial scenarios. Wang et al.’s framework contrasts with ours by focusing on comparing the performance of an improved NSGA-II algorithm adapted to the problem context. Their analysis revealed that their improved NSGA-II algorithm outperformed NSGA-II regarding route length and threat reduction, with modest improvements in concealment. Our framework offers a comprehensive and systematic approach to evaluation. Through multiple experiments across diverse scenarios and specific parameters, it provides a thorough understanding of algorithm performance under various conditions.