Ciencias Exactas y Ciencias de la Salud
Permanent URI for this collectionhttps://hdl.handle.net/11285/551014
Pertenecen a esta colección Tesis y Trabajos de grado de los Doctorados correspondientes a las Escuelas de Ingeniería y Ciencias así como a Medicina y Ciencias de la Salud.
Browse
Search Results
- A hybrid multi-objective optimization approach to neural architecture search for super resolution image restoration(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2025-07) Llano García, Jesús Leopoldo; Monroy Borja, Raúl; emipsanchez; Cantoral Ceballos, José Antonio; Mezura Montes, Efrén; Rosales Pérez, Alejandro; Ochoa Ruiz, Gilberto; School of Engineering and Sciences; Campus Estado de México; Sosa Hernández, Víctor AdriánSuper-resolution image restoration (SRIR) aims to reconstruct a high-resolution image from a degraded low-resolution input. It plays a key role in domains such as surveillance, medical imaging, and content creation. While recent approaches rely on deep neural networks, most architectures remain handcrafted through laborious and error-prone trial-and-error processes. Neural Architecture Search (NAS) seeks to automate the design of deep models, balancing predictive accuracy with constraints like latency and memory usage. Formulating NAS as a bi-level, multi-objective optimization problem highlights these trade-offs and motivates the development of flexible search spaces and strategies that prioritize both performance and efficiency.Prior NAS efforts for SRIR frequently rely on fixed cell structures, scalarized objectives, or computationally intensive pipelines, limiting their practicality on resourceconstrained platforms. Benchmarking shows that such methods often struggle to jointly minimize parameters, FLOPs, and inference time without compromising image reconstruction quality.We propose the Branching Architecture Search Space (BASS), a layer-based, multidepth, multi-branch design that supports dynamic selection, allocation, and repetition of operations. To explore BASS, we introduce a hybrid NAS framework that combines NSGA-III with hill-climbing refinements, guided by SynFlow as a zero-cost trainability estimator. The hybrid approach achieves superior trade-offs in trainability, parameter efficiency, and computational cost when given the same number of function evaluations as vanilla NSGA-III—and reaches comparable Pareto-front approximations with substantially fewer evaluations. The resulting solutions offer enhanced model quality, reduced complexity, and improved deployment suitability for real-world SRIR tasks.Extensive search experiments yield a diverse Pareto front of candidate architectures. Representative designs are fully trained on DIV2K and evaluated across standard SR benchmarks (Set5, Set14, BSD100, Urban100) at →2, →3, and →4 upscales. Balanced models achieve competitive PSNR while operating with significantly fewer parameters and FLOPs than heavyweight baselines. The hybrid search demonstrates faster convergence and improved trade-off resolution compared to single-strategy alternatives, as supported by Bayesian statistical analysis.The combination of BASS and hybrid NSGA-III enables the discovery of SRIR architectures that effectively balance accuracy and resource constraints. This approach facilitates deployment on embedded and real-time systems and offers a generalizable framework for resource-aware NAS across other dense prediction tasks.
- A minutiae-based indexing algorithm for latent palmprints(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12-11) Khodadoust, Javad; Monroy Borja, Raúl; emipsanchez; Aparecida Paulino, Alessandra; Valdes Ramírez, Danilo; Rodríguez Ruiz, Jorge; School of Engineering and Sciences; Campus Monterrey; Medina Pérez, Miguel ÁngelToday, many countries rely on biometric traits for individual authentication, necessitating at least one high-quality sample from each person. However, countries with large populations like China and India, as well as those with high visitor and tourist volumes like France, face challenges such as data storage and database identification. Latent palmprints, comprising about one-third of prints recovered from crime scenes in forensic applications, require inclu sion in law enforcement and forensic databases. Unlike fingerprints, palmprints are larger, and features such as minutiae are approximately ten times more abundant, accompanied by more prominent and wider creases. Consequently, accurately and efficiently identifying la tent palmprints within stored reference palmprints poses significant challenges. Using fre quency domain approaches and deep convolutional neural networks (DCNNs), we present a new palmprint segmentation method in this work that can be used for both latent and full impression prints. The method creates a binary mask. Additionally, we introduce a palmprint quality estimation technique for latent and full impression prints. This method involves parti tioning each palmprint into non-overlapping blocks and considering larger windows centered on each block to derive frequency domain values, effectively accounting for creases and en hancing overall quality mapping. Furthermore, we present a region-growing-based palmprint enhancement approach, starting from high-quality blocks identified through our quality es timation method. Similar to the quality estimation process, this method operates on blocks and windows, transforming high-quality windows into the frequency domain for processing before reverting to the spatial domain, resulting in improved neighboring block outcomes. Finally, we propose two distinct minutiae-based indexing methods and enhance an existing matching-based indexing approach. Our experiments leverage three palmprint datasets, with only one containing latent palmprints, showcasing superior accuracy compared to existing methods
- Botnet detection on twitter: a novel similarity-based clustering mechanism(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12) Samper Escalante, Luis Daniel; Monroy Borja, Raúl; emipsanchez; Castro Espinoza, Félix Agustín; González Mendoza, Miguel; School of Engineering and Sciences; Rectoría Tec de Monterrey; Loyola González, OctavioBotnet detection on Twitter represents a critical yet under-explored research problem,as botnets programmed with malicious intent threaten the platform’s security and credibility. Although Twitter has implemented mitigation strategies, such as imposing restrictions andbans, these measures remain insufficient due to botnets’ rapid creation and expansion. Existing solutions proposed by researchers for manual and automated botnet detection typically rely on individual metrics commonly used for detecting bots. However, these approaches lack the necessary group-oriented analysis and metrics critical for effectively identifying botnets of varying sizes and objectives. To address this issue, we have developed an innovative botnet detection mechanism based on similarity, which significantly enhances the detection rate of botnets on Twitter. Each bot, regardless of its complexity, leaves detectable traces of automation in its creation, behavior, or interactions with other accounts. By characterizing these traces, we can establish relationships between bots, enabling effective botnet detection. Our mechanism constructs a regression model to quantify the similarity between bots, leveraging features from user data, tweet patterns, and social interactions on the platform. Then, it uses this similarity measure to build a distance matrix, enabling the formation of groups with shared attributes, connections, and objectives through clustering methods. Our botnet detection mechanism achieved extraordinary success, evidenced by high scores on external Clustering Validation Indices (CVIs) and the Area under the ROC Curve (AUC) compared to existing solutions from the literature. Furthermore, the mechanism proved effective when confronted with unknown botnets with varied objectives. Our experimental findings suggest that this work is well-positioned to strengthen future botnet detection mechanisms, having shown the value of incorporating social interaction features. This integration offers a strategic advantage in the ongoing arms race against botmasters and their malicious objectives. Additionally, our mechanism consistently outperforms other approaches across various metrics, configurations, and algorithms, underscoring its effectiveness and adaptability in different detection scenarios.
- A novel functional tree for class imbalance problems(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2022-11) Cañete Sifuentes, Leonardo Mauricio; CAÑETE SIFUENTES, LEONARDO MAURICIO; 787723; Monroy Borja, Raúl; puemcuervo, emipsanchez; Morales Manzanares, Eduardo; Gutiérrez Rodríguez, Andrés Eduardo; Cantú Ortiz, Francisco; Conant Pablos, Santiago; School of Engineering and Sciences; Campus Estado de México; Medina Pérez, Miguel AngelDecision trees (DTs) are popular classifiers partly because they provide models that are easy to explain and because they show remarkable performance. To improve the classification performance of individual DTs, researchers have used linear combinations of features in inner nodes (Multivariate Decision Trees), leaf nodes (Model Trees), or both (Functional Trees). Our general objective is to develop a DT using linear feature combinations that outperforms the rest of such DTs in terms of classification performance as measured by the Area Under the ROC Curve (AUC), particularly in class imbalance problems, where one of the classes in the database has few objects compared to another class. We establish that, in terms of classification performance, there exists a hierarchy, where Functional Trees (FTs) surpass Model Trees, that in turn surpass Multivariate Decision Trees. Having shown that Gama's FT, the only FT to date, has the best classification performance, we identify limitations that hinder its classification performance. To improve the classification performance of FTs, we introduce the Functional Tree for class imbalance problems (FT4cip), which takes care in each design decision to improve AUC. The decision of what pruning method to use led us to the design of the AUC-optimizing Cost-Complexity pruning algorithm, a novel pruning algorithm that does not degrade classification performance in class imbalance problems because it optimizes AUC. We show how each design decision taken when building FT4cip contributes to classification performance or to simple tree models. We demonstrate through a set of tests that FT4cip outperforms Gama's FT and excels in class imbalance problems. All our results are supported by a thorough experimental comparison in 110 databases using Bayesian statistical tests.