South Dakota State University
Public Research Access Institutional Repository and Information ExchangeNot a member yet
30309 research outputs found
Sort by
SDSU Data Science Symposium Banquet, 2025
No full texthttps://openprairie.sdstate.edu/ds_symposium_2025_banquet_gallery/1036/thumbnail.jp
Administration Office Records
Full text linkThe Vice President for Administration oversees campus support, service, and administrative units, ensuring efficient operations and strategic direction. Responsibilities include financial management, personnel services, facilities maintenance, student services, legal affairs, public relations, policy development, campus planning, and alumni and foundation support. This role is vital to the university’s stability, growth, and ability to serve students and faculty effectively.
This collection includes a file on South Dakota State University’s 1987-1989 strategic plan, offering insights into institutional priorities and administrative decisions during that period. It serves as a resource for understanding past strategies and their long-term impact. Additionally, the collection contains an open house invitation for Vice President Mike Reger’s retirement, marking a leadership transition and recognizing his contributions to the university
Computation-Efficient Deep Learning Models for Computer Vision and Multimodal Vision-Language Tasks via Network Pruning
No full textWith the rapid evolution of deep neural networks over the past decade, the demand for efficient, generalizable, and task-adaptable models, especially in computer vision, has increased significantly. To address the computational and deployment challenges posed by overparameterized models, the research community has extensively explored model compression techniques such as pruning, quantization, and distillation. These approaches aim to enhance model efficiency without compromising performance, particularly when adapting to domain-specific tasks under limited resources. This dissertation investigates several underexplored yet critical aspects of task-aware deep learning model compression, spanning both convolutional and vision-language architectures. In the early part of this work, we demonstrate that shallow convolutional neural networks, when carefully initialized with orthogonal weights and constrained using orthogonal regularization, can outperform deeper counterparts on specialized classification tasks. Specifically, we show that models tailored to scale-sensitive feature learning can yield competitive performance with significantly fewer parameters. Building on this, we introduce a novel framework for adaptive, structural pruning using deep reinforcement learning. By modeling the pruning decision process as a state-action optimization problem, our agent dynamically adjusts pruning ratios based on the intrinsic dimensionality of training data—a proxy for task complexity. This approach eliminates the need for retraining or extensive post-pruning fine-tuning and reduces compute overhead, offering a more efficient and automated path to model compression. Furthermore, we show that careful reward design and action space construction are pivotal to the agent’s success, particularly when targeting structured modules such as convolutional filters. In the final chapter, we extend the pruning framework to transformer-based vision-language architectures and introduce a task-agnostic, data-dependent approach for structured pruning during continual pretraining. This strategy emphasizes module-aware compression and integrates online knowledge distillation to preserve alignment with pretrained representations. Separately, in earlier chapters, we show that applying pruning after a gradient-informed delay during training and using a hybrid action space significantly improves compression outcomes in convolutional models with minimal loss in generalization. The final contribution focuses on large vision-language models, which are increasingly deployed in zero-shot or few-shot scenarios across tasks such as image-text retrieval, captioning, and classification. We propose TA3DP (Task-Agnostic, Data-Dependent Distillation and Pruning), a pruning framework that integrates online knowledge distillation and module-aware compression to preserve pretrained alignment while adapting to new domains. Our findings demonstrate that distilling from pretrained teachers, rather than fine-tuned ones, yields superior performance, particularly for generative tasks like captioning. This contribution addresses the overlooked problem of catastrophic forgetting during domain-specific fine-tuning under compression. Overall, this dissertation contributes a unified view of task-sensitive model optimization through structured pruning, providing scalable solutions for both CNN-based models and large multimodal architectures. These insights and frameworks lay the foundation for efficient, generalization-preserving model deployment in practical AI systems across diverse domains
One Rung at a Time: Climbing the Ladder Toward Aquatic Connectivity in Eastern South Dakota
No full textRoads and streams are synonymous with their connection to movement, the acquisition of resources, and as pathways to new opportunities. However, the factors that define good roads and ecologically connected streams are rarely without conflict. Road infrastructure at stream-road crossings has the potential to fragment stream networks. Tube culverts, where streams pass under the road through metal pipes, are particularly concerning for stream connectivity. Undersized, aging, or inappropriately installed culverts can develop vertical drops at the outflow due to high velocities and stream bed scouring. The outlet drop can function as a barrier, preventing the upstream movement of fish. Due to the prevalence of tube culverts and their potential to impede passage for small-bodied Great Plains stream fishes, novel cost-effective mitigation strategies and field evaluations are needed to inform barrier inventories and prioritization. In this study, I assessed the in-situ effectiveness of a low-cost Denil-type fish ladder designed to facilitate upstream movement through perched tube culverts. Using a multiple before-after-control-impact (MBACI) design, I deployed passive integrated transponder (PIT) arrays at seven sites to monitor bidirectional movements of non-game, small-bodied stream fishes at impact and control sites before and after ladders were installed at impact locations. Next, I used the pre-ladder installation movement data to evaluate the relationship between Southeastern Aquatic Resource Partnership (SARP) barrier severity estimates and observed passage. I used a combination of mixed-effects logistic regression and multistate models to obtain predicted probability of passage and species-specific transition probability estimates to quantify movement across a gradient of barrier severity. I found that tube culverts acted as barriers to fish movement, with few observed passage events and low estimated passage success and transition probability across sites. The effect of the experimental ladder on upstream movement was temporally variable, species-dependent, and influenced by body size. The data collected during the study is somewhat inconclusive in determining the effectiveness of the ladder in aiding upstream movement in dynamic prairie streams. Moreover, the SARP score was predictive of passage probability, but the mid-range SARP scores did not consistently predict observed passage outcomes, suggesting the protocol may lack sensitivity for small-bodied species. Overall, the result highlights multiple overlapping levels of variability when incorporating biodiverse species, their unique movement capacity, and dynamic structural interaction with stream environments. This emphasizes the need for in-situ passage studies to refine barrier severity assessments and to inform the development of rapidresponse barrier mitigations options, such as the ladder prototype
Dairy and Food Science Student Newsletter, August 8, 2025
No full texthttps://openprairie.sdstate.edu/dairy_student-news/1011/thumbnail.jp
Effects of Different Vitamin E Supplementation on Newborn Lamb Serum Vitamin E Concentrations for the First 28 Days of Life
No full textEvaluation of High-Moisture Ear Corn as a Roughage Source in Finishing Diets Fed to Beef Steers
Full text linkAnimal Science Research and Extension Report 2025 (Complete Report)
Full text linkThe purpose of research at SDSU is to provide reference information that represents the various populations of livestock production. Since the researcher cannot apply treatments to every member of a population, he/she must sample the population. The use of statistics allows the researcher and readers the opportunity to evaluate separation of random occurrences and real biological effects of a treatment
Impact of Parental Involvement on Children’s Mental Health
No full textThis study estimates the causal effect of parental involvement on the mental health of primary school children in Bangladesh. Using an instrumental variables strategy that leverages local parental involvement norms, I find that increased parental involvement significantly reduces cognitive and emotional difficulties. Notably, I find that the protective effect of parental involvement varies by socio-economic context, showing weaker effects for children of mothers who married at an older age and those whose fathers live outside the household, but stronger effects in joint families. Furthermore, I identify two key mechanisms underlying these effects: accommodating parenting style and improved household hygiene practices. A supportive and responsive parenting approach can foster emotional security and self-regulation in children, reducing the risk of anxiety and depression. Improved hygiene practices, meanwhile, lower the burden of illness, which in turn alleviates physical stressors that can impair cognitive and emotional development. This study underscores the importance of parental engagement for child mental health in resource-constrained settings
Synchronization Protocols Utilizing Progesterone Implants Associated with Estrus Expression, Vaginal Microbiome, and Immune Response in Beef Cattle
No full textThis thesis aims to provide a deep understanding of presynchronization strategies and the varying factors affecting their efficacy. Beginning with a literature review of the estrous cycle and the endocrinology driving it. Building on this foundation, the physiological basis for estrous synchronization is described, followed by the advancement of pre-synchronization protocols designed to increase estrus and conception. Further, we will focus on the influence of estrus expression on insemination time and conception rates. Finally, how the immune system interacts with the conception process. In addition, the review goes beyond exogenous endocrine manipulation and dives into the emerging role of the microbial communities of the reproductive tract and their influence on pregnancy and inflammation of the reproductive tract. Together, these interconnected fields underscore the multifactorial nature of fertility and highlight opportunities to improve reproductive management strategies in cattle. Two parallel studies were conducted to address the estrous expression and the microbiome related to synchronization protocols that utilize a 14-day implant. The first study focused on characterizing the onset of estrus in heifers, primiparous, and multiparous cows that are synchronized using the 7&7 synchronization protocol and variations. The second characterized the vaginal microbiome changes from the start to the end of the 7&7 synchronization protocol by utilizing 16S rRNA gene sequencing. The understanding of biological and management factors and their contribution to fertility is vital in increasing conception and reducing reproductive loss in cattle