24067 research outputs found
Sort by
GRM-0254 Unified Robust Optimal Transport For Outlier-resilient Learning
No full textClassical Optimal Transport (OT) is particularly sensitive to outliers. The existing robust variant, ROBOT, mitigates this through hard truncation, but its rigidity often compromises stability. We propose WROT-r, a unified r-power framework for weighted robust OT that combines rigorous hard-clipping and smooth cost compression through a single parameter r. WROT-r offers a continuous robustness spectrum, enabling adaptive control over how strongly transport costs are down-weighted for outliers. Experiments on synthetic mean estimation and resilient GANs show clear patterns: larger r performs best under weak contamination by preserving more inliers, while smaller r (≈1.5) is more effective under moderate and strong contamination. The extreme r→1 limit (ROBOT) remains consistently unstable. Overall, WROT-r improves robustness across a broad range of noise conditions
UR-0225 Carbonyl Detection in IR Using Deep Learning
No full textThe goal of this project is to train a Convolutional Neural Network (CNN) to recognize carbonyl groups in infrared (IR) spectra. A carbonyl group is defined by a characteristic C=O double bond, which produces a strong, easily recognizable absorption peak near 1700 cm⁻¹. To develop and evaluate the model, I am using spectra prepared through three common techniques: KBr disc, nujol mull, and liquid film. Among these, liquid-film spectra provide the cleanest signal and most closely resemble what a chemist visually relies on when identifying carbonyls. In contrast, both the KBr disc and nujol mull methods require mixing the target compound with additional materials, which can introduce interference and make automated detection more challenging
KSU Jazz Ensemble II & III
No full textKSU Jazz Ensemble II
Wes Funderburk, Conductor
KSU Jazz Ensemble III
Rob Opitz, Conductorhttps://digitalcommons.kennesaw.edu/musicprograms/2960/thumbnail.jp
Microfluidic Device for Infection Testing
No full textThis research focuses on the early detection of viral infections. We propose a method that utilizes magnetically tagged antigens, which are transported through a microfluidic solenoid channel by a neodymium magnet. The induced voltage signal, measured at the nanovolt scale, presents the indication of viral presence, potentially allowing for detection prior to the conventional testing methods. Building upon previous studies, our work simulates the dynamic motion of the magnetic beads by integrating magnetic fields with structural mechanics, resulting in more precise outcomes. We use COMSOL Multiphysics to conduct these simulations, demonstrating the feasibility and effectiveness of our proposed approach. In addition to the proposed magnet, the redesigned circuit is an iteration that turns our previous wide-band dual-stage circuit into a precise, three-stage active low pass amplifier. Each LTC2050 stage provides a gain of -100 with 0.39 microfarad capacitors, forming cascaded first-order filters with an overall -3 dB cutoff near 20 Hz and overall gain of
106106
volts/volt. These improvements produced higher signal to noise ratio and enhanced output amplitudes, eliminating the need for digital signal processing. To manufacture the solenoid, the winding apparatus we developed automated the process of winding a 50 nanometer thick copper wire around a 170 micrometer thick optical fiber, a precise task that could not be done manually. We achieved the winding of the optical fiber using 2 stepper motors: one rotated the spindle holding the fiber, the other incrementally translated the stage after each full rotation to control coil pitch and direction. This coordinated sequence repeated for 25 full iterations, which produced a uniform, tightly wrapped coil along the fiber’s surface. The automation ensured repeatable, high-quality winding suitable for micro-fluidic detection. Additionally, AI was used in the creation of this abstract to verify and correct writing errors as well as fact-check information
Sigma-1 Receptor Mediated Modulation of Glutamate Receptors Enhances NMDA Channel Activity and Memory Processes
No full textGlutamate receptors are key mediators of excitatory neurotransmission and are essential for synaptic plasticity, the cellular foundation of learning and memory. Among these, AMPA receptors mediate rapid excitatory signaling, whereas NMDA receptors are critical for synaptic strengthening and long-term plasticity. Precise regulation of NMDA receptor (NMDAR) function is vital for maintaining neuronal communication and cognitive stability. The sigma-1 receptor (σ1R), an endoplasmic reticulum chaperone localized at mitochondria-associated membranes, has been shown to influence glutamatergic transmission and cognitive processes. Activation of σ1R enhances glutamate release and promotes long-term potentiation (LTP)—a key cellular correlate of memory formation—in hippocampal neurons, thereby supporting the molecular mechanisms of learning and memory. However, the direct impact of σ1R on NMDAR channel activity remains incompletely understood. We hypothesize that σ1R acts as a positive modulator of NMDAR function, enhancing receptor-mediated synaptic signaling and contributing to memory processes. To test this, electrophysiological recordings were conducted on hippocampal synaptosomes to characterize NMDAR channel dynamics following σ1R activation. Treatment with a σ1R agonist significantly increased both the mean open probability and conductance of NMDAR channels, indicating potentiation of receptor activity. This enhancement correlates with improved synaptic plasticity and cognitive performance. Collectively, these findings identify σ1R as a critical modulator linking glutamatergic neurotransmission with memory formation and suggest that σ1R activation represents a promising therapeutic strategy for mitigating cognitive decline and neurodegenerative disorders
Playa Lakes Play a Vital Role in the High Plains Ecosystem
No full textThe High Plains Region of the United States, extending from Southern South Dakota to West Texas, is home to vast grasslands and playa lakes—shallow depressions that form temporary wetlands. These playa lakes are vital for recharging the High Plains Aquifer, also known as the Ogallala Aquifer, one of the largest sources of freshwater in the country. Despite limited rainfall, this region has historically supported rich grassland ecosystems and remains crucial for both global agricultural production and migratory bird populations traveling the Central Flyway.
My poster explores the ongoing depletion of the playa lakes and grasslands due to over-irrigation and land conversion, an often-overlooked issue with wide-reaching environmental consequences. These changes threaten biodiversity, disrupt the water cycle and pose a threat to long term agricultural production. The aim of my project is to raise awareness among the general public, who are often unaware of the critical ecological role of the High Plains Region.
One of the biggest challenges I faced during this project was locating usable shapefiles to visually represent grassland to cropland conversion. My initial attempts to show grassland conversion were complicated by overly detailed cropland data. Eventually, I shifted focus after finding an irrigation shapefile which showed a much clearer visual of the use of water in the area. The other part of the story is told by the Greater Prairie Chicken, the Lark Bunting and the Cassin’s Sparrow. The three species of birds rely greatly on the playa lakes as a means for survival throughout breeding and migratory journeys.
With personal ties to the southern High Plains, I hope this poster brings attention to the ecological importance of this region and inspires more informed land-use decisions
Functional Analysis of MKNK2 Isoforms in response to MAPK phosphorylation
No full textMKNK2 is a serine/threonine kinase activated downstream of the mitogen-activated protein kinases (MAPK) pathway. p38 and ERK1/2 are known kinase activators of MKNK2, though MKNK2 is known to preferentially interact with ERK1/2. MKNK2 belongs to the MKNK family, which includes two genes, MKNK1 and MKNK2. These kinases phosphorylate the eukaryotic initiation factor eIF4E, which is a regulator of mRNA translation, cell growth, and oncogenic signaling. In recent years, MKNK2 has been studied for its role in cancer, where phosphorylation of eIF4E promotes tumorigenesis. MKNK2 is alternatively spliced into two isoforms, MKNK2-long and MKNK2-short. The long form has been reported to act as a tumor suppressor, while the short form functions as an oncogene. In this project, we purified GST-versions of both MKNK2 proteins and performed a kinase assay to test the ability of both isoforms to be a substrate of ERK or p38 MAP kinase, and if the short and long isoforms can phosphorylate kemptide after activation. The information from these experiments could help with the development of targeted inhibitors of oncogenic pathways
Adaptive Refuge: Reimagining Biophilic Settlements for Sudanese Refugees in Addis Ababa, Ethiopia
No full textEvery day, thousands of people are forced to flee their homes escaping war, disaster, and instability. What begins as a temporary refuge often becomes a lifetime condition. In Ethiopia, over 4.6 million refugees live within this state of protracted displacement, with more than one-fourth arriving from Sudan after the 2023 conflict. Most remain in remote, rural camps along border regions far from the educational, economic, and social opportunities of the city. These spatial and social divides reinforce dependency, limit access to livelihood, and erode personal dignity. Yet proximity to urban infrastructure could offer a pathway toward belonging, participation, and long-term resilience.This thesis asks: How can refugee settlements be re-envisioned as adaptive, biophilic, and semi-permanent communities that promote healing, education, and livelihood? Centered in Akaki–Kality, a transitional district on the southern edge of Addis Ababa,Ethiopia the project proposes a new model for urban-accessible refugee living. Through biophilic design principles, it reframes the refugee camp as a living system composed of adaptive housing clusters, wellness and learning centers, and shaded courtyards that connect people through ecology and community. Using precedent analysis and qualitative research, the study investigates how architecture can merge environmental systems, material adaptability, and human-centered design to create places of belonging rather than containment. Ultimately, this project envisions displacement not as isolation, but as integration where architecture becomes a catalyst for ecological restoration, social empowerment, and collective healing. It positions design as a bridge between humanitarian relief and long-term urban development, transforming the refugee camp from a symbol of crisis into a framework for regeneration.
Keywords: refugee architecture, adaptive design community integration, urban accessibility, displacement, biophilic design
*Portions of this abstract were developed with the assistance of an artificial intelligence (AI) tool, which was used solely to outline and organize the structure of the text. All final content, revisions, and ideas were written, verified, and approved by the author.
Autumn Lite LLM
No full textAutumn Lite is an inspectable, small-footprint language modeling pipeline for reproducible experimentation and practical integration into video-game non-player character (NPC) systems. It comprises four components: (1) a regex-aware tokenizer/normalizer for vocabulary construction and mixed prose–code handling; (2) a classical evaluation track that reports perplexity to quantify predictive quality; (3) a compact neural language model (decoder-only Transformer) targeted at low latency and controllable outputs; and (4) a lightweight sentiment classifier (logistic regression) that assigns positive/neutral/negative tags to steer text-to-speech (TTS) prosody during NPC dialogue. By combining transparent preprocessing with baseline metrics and a small, deployable decoder, Autumn Lite aims to deliver predictable, designer-friendly behavior for NPC speech, enabling subtle, real-time adjustments to rate, pitch, and emphasis instead of monotone delivery.
“This system operates as a standard small LLM and can be combined with NPC dialogue/TTS; in this presentation I will cover only the LLM portion.
Understanding the Role of hlh-14 and ceh-27 in the M4 Motor Neuron in C. elegans
No full textAttention Deficit Hyperactivity Disorder (ADHD) affects millions worldwide and can shape social interactions, health factors, and overall quality of life. The genetic mechanisms underlying ADHD, however, are not well defined. One gene linked to ADHD, Nkx2.1, plays a key role in neuronal differentiation across mammals. Because these transcriptional networks are deeply conserved, studying Nkx2.1’s ortholog, ceh-27, in C. elegans allows these mechanisms to be examined in a simpler, well-mapped nervous system. The worm’s invariant cell lineage and genetic accessibility make it an ideal model for discovering conserved neurodevelopmental pathways. This study investigates the regulatory relationship between ceh-27 and hlh-14, a basic helix-loop-helix transcription factor that promotes early neurogenesis, in the M4 motor neuron. Using C. elegans strains expressing hlh-14::GFP, fluorescence microscopy will compare hlh-14 expression in wild-type and ceh-27 mutant embryos. Spatial and intensity analyses will determine whether ceh-27 represses or activates hlh-14 during M4 specification. By defining how ceh-27 interacts with hlh-14 in a single, invariant neuron (M4), this research will provide more insight into the transcriptional hierarchies that govern neuronal identity in C. elegans and conserved mechanisms relevant to neurodevelopmental disorders