Theses & Dissertations
Permanent URI for this collection
This collection is exclusively made up of theses, dissertations, and project reports submitted as a requirement for completing a graduate degree at UPR-Mayagüez.
If you are a UPRM graduate student and you are looking for information related to the deposit process, please refer to https://libguides.uprm.edu/repositorioUPRM/tesis
Browse
Recent Submissions
1 - 5 of 2998
-
PublicationA Bayesian machine learning approach for EEG functional connectivity estimation and working memory load classification in human subjects( 2024-05)Analyzing working memory is essential for understanding cognitive processes and improving educational strategies, mental health diagnostics, and psychological interventions. Electroencephalogram (EEG) signals, known for their high temporal correlation, effectively capture these subtle responses, highlighting the importance of assessing EEG-based functional connectivity across various frequency bands to understand brain dynamics under varying cognitive loads. Traditional methods, typically involving regression models, often face challenges like biased connectivity estimates due to enforced sparsity and inaccuracies from small sample sizes or sampling noise. Addressing these issues, the current study develops a Bayesian structure learning algorithm to learn the functional connectivity of EEG. This approach ensures accurate connectivity analyses across different frequency bands. Next, functional connectivity features are given as an input to graph convolutional network to classify working memory loads. This study analyzes five working memory datasets to evaluate the proposed methodology. The subject-specific classification yields an average sensitivity and specificity of 92% and 94%, respectively. The proposed methodology produced consistent results in functional connectivity estimation compared to state-of-the-art functional connectivity metrics. The study finds that encoding information is critical in altering functional connectivity for different working memory loads rather than its manipulation/retention of tasks.
-
PublicationMeta-analysis for crop fertility studies in Puerto Rico using linear mixed models and nonlinear mixed models( 2024-05-09)A popular technique for increasing crop yields worldwide is nitrogen fertilization. However, excessive nitrogen fertilizer causes enormous emissions of greenhouse gases, which contribute to global warming and climate change. In addition, unused nitrogen contaminate water causing problems to aquatic life. Having nutrient management in terms of fertilizer recommendations is key to having a sustainable high yielding crop, without damaging the plant, environment, and overall soil productivity. This project’s application consists in estimating the total amount of fertilizer, nitrogen, needed by a crop using statistical modeling. The relationship between variables for crop and soil processes are often better captured by nonlinear models because they provide a series of advantages. In this project, the crop nutrient requirement (CNR) is used as a non typical effect size for fertilizer recommendations. This metric depends on the fertilizer rate and the crop’s relative yield. Using these two metrics, crop yield response curves can be obtained from different agricultural studies. In order to get a combined CNR estimate this project will consider using two different meta-analysis methodologies: Linear Mixed Models (LMMs), and Non-Linear Mixed Models (NLMMs). Field fertility research conducted with crops from the Solanaceae family and forage crops documenting yield response to nitrogen fertilizers were used for this project. The LMM methodology consists on a two step approach: (1) the exponential model was used in order to get CNR estimates for all studies, and (2) these results were used as observation to fit an LMM and obtain the general CNR estimate. On the other hand, the second approach consisted in fitting an exponential NLMM to the raw data in order to obtain a different combined CNR estimate. Results show that 95% confidence intervals for the CNR combined estimates after meta-analysis were narrower than the individual confidence intervals from each study.
-
PublicationRetos y necesidades de los caficultores y caficultoras de café especial para mantener y mejorar su producción en Puerto Rico( 2024-05-09)The purpose of the research was to establish the challenges and needs of specialty coffee farmers to maintain and improve their production in Puerto Rico, and whether producing specialty coffee results in benefits. The population was composed of 20 coffee growers whose coffees scored above 80 points (specialty coffee) in cupping. A six-part questionnaire with a total of 41 questions was used for data collection. The most important challenge was the prolonged drought, and the greatest need for training was the prevention and management of nutritional deficiencies in the coffee plant and its effect on the cup. This research validated that for coffee growers, producing specialty coffee motivated them to continue in the coffee industry, obtained a better price when selling their coffee and improved the quality of life of them and their families.
-
PublicationEfectos de la disponibilidad de espacio en la salud y en la eficiencia en conversión de alimento para el aumento en peso y desarrollo esqueletal en becerros Holstein( 2024-05-09)The objective of this research was to analyze whether providing more space to calves could improve their growth, body condition, health status, milk and feed consumption, feed conversion efficiency and FARM evaluation from birth to weaning. Holstein calves (n=13) were evaluated for 10 weeks under an accelerated growth feeding protocol in conventional 0.76m2 cages (control=CON) or large 1.95m2cages (AMP). The hypothesis of this research was that providing more space will improve body condition, health, growth and feed conversion efficiency, and better scores will be obtained in FARM evaluations in Holstein calves. An evaluation of calf health status, growth, body condition (BCS) and FARM evaluation was carried out weekly. In the investigation, no difference was found in weight (P=0.8570) or average daily gain (P=0.9173) between treatments. In conclusion, differences were only obtained in locomotion, also the study showed a saving in the cost of weight gain of $0.57 per kg when using the AMP cages.
-
PublicationDeep merge-and-run convolutional neural network for image denoising and super-resolution( 2024-05-10)Imaging systems have become part of ordinary life due to their integration into smart devices such as phones, tablets, and computers. In the field of image processing, we face a common challenge: noisy and low-resolution images. There are limitations and inherent problems to optical and imaging systems. Some of these are caused by noise, lighting, or vibrations. Recent years have seen significant advances in deep learning and machine learning techniques have been aimed at addressing these complex problems. This thesis introduces an innovative deep learning architecture composed of convolutional blocks in a merge-and-run configuration. The model tackles two prevalent problems in image processing: image denoising and single-image super-resolution. Our focus includes the design, implementation, and evaluation of this architecture, specifically targeting the denoising of confocal microscopy images and super-resolution of synthetic aperture radar (SAR) images. The model achieved highly competitive results in both use cases, enhancing image clarity and resolution comparably to existing methods, but with a 45% reduction in architecture size.