BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Events - ECPv6.16.5.1//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-ORIGINAL-URL:https://events.ucsc.edu
X-WR-CALDESC:Events for Events
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:20240310T100000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:20241103T090000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:20250309T100000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:20251102T090000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:20260308T100000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:20261101T090000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20250915T100000
DTEND;TZID=America/Los_Angeles:20250915T100000
DTSTAMP:20250925T231435Z
CREATED:20250905T070000Z
LAST-MODIFIED:20250925T231435Z
UID:10000145-1757930400-1757930400@events.ucsc.edu
SUMMARY:Van Duker\, N. (AM) - A Random Choice Hybrid Method for Resolving Shock Placement Errors in 1D Relativistic Hydrodynamics with Transverse Velocities
DESCRIPTION:This report presents a one-dimensional Random Choice-based hybrid method for simulating special relativistic hydrodynamics (SRHD) flow problems. The proposed scheme combines a high-order accurate method and a random choice method\, selectively applying the first to smooth flows and the second to shocks and discontinuities. This hybrid approach addresses the issue of incorrect wave placements in the presence of significant transverse velocity\, commonly encountered in one-dimensional SRHD shock tube tests. In support of this development\, we present a modified shock/contact detection switch\, specifically tuned for relativistic flows. We find that our method improves both the accuracy and computational performance when compared against existing methods on a well-known family of pathological shock tube problems. Our analysis of these pathological problems provides a path forward for further improving existing higher-dimensional methods. \nEvent Host: Nathan Van Duker\, Ph.D Student\, Applied Mathematics \nAdvisor: Dongwook Lee
URL:https://events.ucsc.edu/event/van-duker-n-am-a-random-choice-hybrid-method-for-resolving-shock-placement-errors-in-1d-relativistic-hydrodynamics-with-transverse-velocities/
LOCATION:Engineering 2\, Engineering 2 1156 High Street\, Santa Cruz\, CA\, 95064
GEO:37.0009723;-122.0632371
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Engineering 2 Engineering 2 1156 High Street Santa Cruz CA 95064;X-APPLE-RADIUS=500;X-TITLE=Engineering 2 1156 High Street:geo:-122.0632371,37.0009723
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20250910T140000
DTEND;TZID=America/Los_Angeles:20250910T140000
DTSTAMP:20250925T231632Z
CREATED:20250825T070000Z
LAST-MODIFIED:20250925T231632Z
UID:10000123-1757512800-1757512800@events.ucsc.edu
SUMMARY:Mavrogiannakis\, A. (CSE) - Scalable Oblivious Databases and Systems
DESCRIPTION:Modern applications are increasingly designed with a strong emphasis on scalability and performance\, as systems are expected to process ever-growing volumes of data and deliver results with minimal latency. Techniques such as distributed architectures\, in-memory computation\, and optimized data structures are routinely adopted to meet these performance-driven demands. However\, in the pursuit of speed and efficiency\, security is often treated as a secondary concern or an afterthought. This oversight can lead to critical vulnerabilities\, as even the most performant systems remain fundamentally insecure if sensitive information can be leaked or exploited. As data becomes more valuable and privacy regulations grow stricter\, ensuring robust security measures is not merely desirable but strictly necessary—an essential requirement that must stand alongside scalability and performance as a first-class design goal. \nTo meet security requirements\, many applications adopt end-to-end encryption to protect data stored in the cloud. While this prevents external adversaries from accessing sensitive information\, prior work [CITE] has demonstrated that encryption alone is insufficient: an untrusted server can still exploit execution patterns and access behaviors to gradually reconstruct the underlying database in plaintext. As an alternative\, other applications rely on Trusted Execution Environments (TEEs)\, which offer strong guarantees through memory encryption\, isolation\, and integrity checks. TEEs are particularly appealing due to their ease of use and high performance\, often approaching that of non-encrypted systems. However\, TEEs are not without limitations [CITE]. They remain vulnerable to leakage-abuse attacks and side-channel vulnerabilities [CITE]\, which can undermine their security guarantees in practice. \nIn my research\, I combine TEEs with oblivious computation to achieve stronger security guarantees without sacrificing practicality. Specifically\, my work focuses on designing\, analyzing\, and implementing oblivious algorithms for databases and systems. A central theme of my research is bridging the gap between security and performance\, developing scalable algorithms that approach the efficiency of plaintext execution. For example\, in our first project\, Obliviator (to appear at USENIX Security ’25)\, we introduced oblivious implementations of fundamental database operators—such as filtering\, aggregation\, and joins—in a shared-memory setting\, achieving efficiency at scale on datasets up to hundreds of gigabytes. Building on this foundation\, our subsequent work extends these operators to distributed environments\, addressing challenges such as secure execution under weaker trust assumptions and reducing communication overhead\, both in terms of rounds and data exchanged. We also introduced frameworks that enable parallelism in oblivious computation\, further enhancing performance. My current work focuses on extending these techniques to multi-way joins\, where combining multiple tables introduces new challenges in both efficiency and security. In parallel\, I am exploring query optimization strategies tailored to the oblivious setting\, with the goal of pushing oblivious database systems closer to the performance of traditional plaintext systems. \nEvent Host: Apostolos Mavrogiannakis\, Ph.D Student\, Computer Science & Engineering
URL:https://events.ucsc.edu/event/mavrogiannakis-a-cse-scalable-oblivious-databases-and-systems/
LOCATION:Engineering 2\, Engineering 2 1156 High Street\, Santa Cruz\, CA\, 95064
GEO:37.0009723;-122.0632371
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Engineering 2 Engineering 2 1156 High Street Santa Cruz CA 95064;X-APPLE-RADIUS=500;X-TITLE=Engineering 2 1156 High Street:geo:-122.0632371,37.0009723
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20250827T110000
DTEND;TZID=America/Los_Angeles:20250827T110000
DTSTAMP:20250925T231632Z
CREATED:20250825T070000Z
LAST-MODIFIED:20250925T231632Z
UID:10000124-1756292400-1756292400@events.ucsc.edu
SUMMARY:Rakshit\, G. (CSE) -Improving Question Answering through Figurativeness Understanding\, Semantic Representation and Multi-Agent Conflict Resolution
DESCRIPTION:Open-domain question answering (ODQA) systems come with diverse challenges — ranging from resolving conflicting information to interpreting figurative expressions and representing meaning in a human-understandable form. This dissertation presents three complementary contributions toward building more robust and interpretable QA systems. \nFirst\, we investigate QA model performance on figurative language. Introducing FigurativeQA\, a benchmark of yes/no questions with figurative and literal contexts\, we demonstrate that popular BERT-based QA systems underperform significantly on figurative text. However\, prompting-based approaches like ChatGPT with chain-of-thought reasoning can mitigate this gap\, particularly when figurative contexts are automatically simplified. \nSecond\, we present ASQ\, a novel tool for automatically generating question-answer meaning representations (QMR) from Abstract Meaning Representation (AMR) graphs. ASQ enables scalable and linguistically grounded QA dataset construction\, bridging traditional formal semantics with natural language interfaces. We show that ASQ-generated questions exhibit high content fidelity and overlap with existing crowd-annotated resources like QAMR. \nFinally\, we explore how large language models (LLMs) handle conflicting evidence in ODQA\, proposing a multi-agent framework where answers generated by different models are evaluated through a verification step. Experiments using the QACC dataset and state-of-the-art LLMs (GPT-4o\, Claude 4\, DeepSeek-R1) reveal that model diversity enhances answer quality\, though requiring explanations during verification does not always lead to improvements. \nTogether\, these contributions advance the interpretability\, robustness\, and accuracy of QA systems. \nEvent Host: Geetanjali Rakshit\, Ph.D Candidate\, Computer Science & Engineering
URL:https://events.ucsc.edu/event/rakshit-g-cse-improving-question-answering-through-figurativeness-understanding-semantic-representation-and-multi-agent-conflict-resolution/
LOCATION:Engineering 2\, Engineering 2 1156 High Street\, Santa Cruz\, CA\, 95064
GEO:37.0009723;-122.0632371
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Engineering 2 Engineering 2 1156 High Street Santa Cruz CA 95064;X-APPLE-RADIUS=500;X-TITLE=Engineering 2 1156 High Street:geo:-122.0632371,37.0009723
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20250821T130000
DTEND;TZID=America/Los_Angeles:20250821T130000
DTSTAMP:20250925T231627Z
CREATED:20250815T070000Z
LAST-MODIFIED:20250925T231627Z
UID:10000106-1755781200-1755781200@events.ucsc.edu
SUMMARY:Briden\, M. (CSE) -  Representation Learning and Generative Forecasting for Noisy and Limited Clinical Data: Applications in Wound Healing and EEG
DESCRIPTION:The rapid integration of artificial intelligence and machine learning into clinical practice has driven advances in disease classification\, segmentation\, and clinical decision support. However\, the complexities of medical data pose a challenge to widespread adoption. The rarity of medical conditions\, ethical considerations\, and varying acquisition protocols leads to limited and noisy data. The time-intensive process of labeling data\, the high degree of accuracy required in clinical settings\, and the ill-defined nature of certain medical conditions further complicate the application and deployment of machine learning models. Likewise\, high‐stakes medical decisions demand trustworthy and interpretable predictions. However\, prioritizing trust and explainability is rarely a primary objective in most model designs. \nThis thesis addresses three key challenges in machine learning for healthcare. First\, we develop methods for learning under noisy and limited medical data\, focusing on representation learning strategies that improve generalization when datasets are small or contain mislabeled samples. Second\, we explore the prediction of generative outcomes amid label noise and data scarcity\, utilizing parameter-efficient and temporal generative models to forecast disease trajectories. Third\, we advance trustworthy and explainable medical artificial intelligence by designing deep architectures that provide interpretable outputs suitable for clinical decision-making. \nThese challenges are addressed in the context of two complementary medical modalities: wound healing images and electroencephalogram signals. Wound healing tasks focus on predicting healing trajectories while enhancing interpretability through segmentation-based explanations and training large models in light of extreme data noise and scarcity. Electroencephalogram-based tasks emphasize representation learning and explainability for non-invasive mental state classification. These experiments demonstrate the clinical relevance of the proposed approaches and their ability to operate under challenging medical conditions across both imaging and physiological signal domains. \nEvent Host: Michael Briden\, PhD Candidate\, Computer Science & Engineering \nAdvisor: Narges Norouzi
URL:https://events.ucsc.edu/event/briden-m-cse-representation-learning-and-generative-forecasting-for-noisy-and-limited-clinical-data-applications-in-wound-healing-and-eeg/
LOCATION:Engineering 2\, Engineering 2 1156 High Street\, Santa Cruz\, CA\, 95064
GEO:37.0009723;-122.0632371
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Engineering 2 Engineering 2 1156 High Street Santa Cruz CA 95064;X-APPLE-RADIUS=500;X-TITLE=Engineering 2 1156 High Street:geo:-122.0632371,37.0009723
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20250819T140000
DTEND;TZID=America/Los_Angeles:20250819T140000
DTSTAMP:20250925T231432Z
CREATED:20250818T070000Z
LAST-MODIFIED:20250925T231432Z
UID:10000111-1755612000-1755612000@events.ucsc.edu
SUMMARY:Bhatia\, N. (CSE) - Building Adaptive Intelligence into Wireless Sensing
DESCRIPTION:WiFi-based indoor positioning is a widely researched area focused on determining the location of devices. Accurate indoor positioning has numerous applications\, including asset tracking and indoor navigation. Despite advances\, their adoption in practice remains limited due to several challenges such as environmental changes that cause signal fading\, multipath effects\, and interference\, all of which reduce positioning accuracy. Moreover\, telemetry data vary across WiFi device vendors\, presenting distinct features and formats\, while use-case requirements can also differ significantly. At present\, there is no unified model capable of handling these variations effectively. \nWe present WiFiGPT\, a decoder-only transformer-based system designed to address these variations while achieving high localization accuracy. Our experiments with WiFiGPT show that it can effectively capture subtle spatial patterns in noisy wireless telemetry\, making them reliable regressors. Compared to state-of-the-art methods\, our approach matches and often surpasses conventional techniques across multiple types of telemetry. Achieving sub-meter accuracy for RSSI and FTM and centimeter-level precision for CSI highlights the potential of LLM-based localization to outperform specialized methods\, without the need for handcrafted signal processing or calibration. Other work includes EchoSense\, which utilizes CSI to monitor vital signs such as heart rate and respiration with high accuracy. \nEvent Host: Nayan Bhatia\, PhD Student\, Computer Science & Engineering \nAdvisor: Katia Obraczka
URL:https://events.ucsc.edu/event/bhatia-n-cse-building-adaptive-intelligence-into-wireless-sensing/
LOCATION:Engineering 2\, Engineering 2 1156 High Street\, Santa Cruz\, CA\, 95064
GEO:37.0009723;-122.0632371
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Engineering 2 Engineering 2 1156 High Street Santa Cruz CA 95064;X-APPLE-RADIUS=500;X-TITLE=Engineering 2 1156 High Street:geo:-122.0632371,37.0009723
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20250819T140000
DTEND;TZID=America/Los_Angeles:20250819T140000
DTSTAMP:20250925T231627Z
CREATED:20250818T070000Z
LAST-MODIFIED:20250925T231627Z
UID:10000112-1755612000-1755612000@events.ucsc.edu
SUMMARY:Swaby\, A. (ECE) -  Improving X-ray Medical Imaging using Amorphous Selenium as a Photoconductive Layer
DESCRIPTION:The presence of coronary artery calcification is a strong predictor for future cardiovascular events where cardiac risk categories are quantified depending on calcification size. Dual-energy chest X-rays provide high contrast visualization to improve opportunistic screening for quantifying coronary artery calcifications\, determining bone mineral density (i.e.\, osteoporosis) and characterizing lung lesions. As a dual-energy imaging modality\, multilayer flat panel detectors acquire low- and high-energy X-ray images as a polyenergetic\, single-exposure. Combining two detectors into a dual-layer configuration\, weighted subtraction techniques in the resulting images allow for differentiation of soft tissue from the projection of the bone structures and other high attenuating materials. To improve detection of calcifications < 1 mm in size\, the performance of a dual-layer X-ray detector is investigated as a means of providing the necessary μm-resolution and spectral separation for enhanced contrast between low- and high-energy X-ray images. A cascaded linear systems model is used to simulate the modulation transfer function\, detective quantum efficiency\, and noise power spectrum of an amorphous selenium direct conversion top detector and a cesium iodide-based indirect conversion bottom detector. As the framework for system design and optimization\, a generalized task-based analysis is used to analyze how the signal projections\, noise contributions\, task function\, and weighting factors contribute to the detectability index of the dual-layer imaging system.  \nEvent Host: Akyl Swaby\, PhD Candidate\, Electrical & Computer Engineering \nAdvisor:  Dr. Shiva Abbaszadeh
URL:https://events.ucsc.edu/event/swaby-a-ece-improving-x-ray-medical-imaging-using-amorphous-selenium-as-a-photoconductive-layer/
LOCATION:Engineering 2\, Engineering 2 1156 High Street\, Santa Cruz\, CA\, 95064
GEO:37.0009723;-122.0632371
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Engineering 2 Engineering 2 1156 High Street Santa Cruz CA 95064;X-APPLE-RADIUS=500;X-TITLE=Engineering 2 1156 High Street:geo:-122.0632371,37.0009723
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20250819T100000
DTEND;TZID=America/Los_Angeles:20250819T100000
DTSTAMP:20250925T231431Z
CREATED:20250815T070000Z
LAST-MODIFIED:20250925T231431Z
UID:10000105-1755597600-1755597600@events.ucsc.edu
SUMMARY:Osorio\, S. (AM) - Image-Based Wound Infection Classification
DESCRIPTION:This thesis investigates the use of deep learning for classifying wound infections from photographic images\, using colony-forming unit (CFU) counts as a quantitative labeling standard. Leveraging the visual information in wound photographs and the clinical relevance of bacterial burden\, the study implements a multi-task U-Net architecture for both image reconstruction and binary classification in a shared-encoder framework. Three experimental conditions were explored: one using original images with positive class weighting\, one incorporating data augmentation to enhance visual diversity\, and one employing 5-fold cross-validation with augmentation to improve validation reliability. The non-augmented model achieved 91.7% accuracy at a threshold of 0.8\, correctly identifying 4 of 5 infected cases\, while Experiment 2 achieved 87.5% accuracy at a moderate threshold of 0.5 but became more conservative at higher thresholds. The third experiment reached 79.6% accuracy at a threshold of 0.3\, detecting all 11 infected cases despite signs of overfitting. These results highlight the model's strong performance in minimizing false negatives\, particularly in the non-augmented setting\, but also reflect limitations from the small dataset\, class imbalance\, and reliance on a small validation set. These factors suggest results should be interpreted cautiously and motivate further study with larger datasets\, improved regularization\, and more varied clinical scenarios. \nEvent Host: Sebastian Osorio\, M.S. Candidate\, Scientific Computing & Applied Mathematics \nAdvsior: Marcella Gomez
URL:https://events.ucsc.edu/event/osorio-s-am-image-based-wound-infection-classification/
LOCATION:Engineering 2\, Engineering 2 1156 High Street\, Santa Cruz\, CA\, 95064
GEO:37.0009723;-122.0632371
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Engineering 2 Engineering 2 1156 High Street Santa Cruz CA 95064;X-APPLE-RADIUS=500;X-TITLE=Engineering 2 1156 High Street:geo:-122.0632371,37.0009723
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20250813T100000
DTEND;TZID=America/Los_Angeles:20250813T100000
DTSTAMP:20250925T231625Z
CREATED:20250806T070000Z
LAST-MODIFIED:20250925T231625Z
UID:10000092-1755079200-1755079200@events.ucsc.edu
SUMMARY:Asefi\, N. (ECE) - Generative Lagrangian Data Assimilation for Ocean Dynamics under Extreme Sparsity
DESCRIPTION:Reconstructing ocean dynamics from observational data is fundamentally limited by the sparse\, irregular\, and Lagrangian nature of spatial sampling\, particularly in subsurface and remote regions. This sparsity poses significant challenges for forecasting key phenomena such as eddy shedding and rogue waves. Traditional data assimilation methods and deep learning models often struggle to recover mesoscale turbulence under such constraints. We leverage a deep learning framework that combines neural operators with denoising diffusion probabilistic models (DDPMs) to reconstruct high-resolution ocean states from extremely sparse Lagrangian observations. By conditioning the generative model on neural operator outputs\, the framework accurately captures small-scale\, high-wavenumber dynamics even at $99%$ sparsity (for synthetic data) and $99.9%$ sparsity (for real satellite observations). We validate our method on benchmark systems\, synthetic float observations\, and real satellite data\, demonstrating robust performance under severe spatial sampling limitations as compared to other deep learning baselines. \nEvent Host: Niloofar Asefi\, PhD Student\, Electrical & Computer Engineering \nAdvisor: Ashesh Chattopadhyay
URL:https://events.ucsc.edu/event/asefi-n-ece-generative-lagrangian-data-assimilation-for-ocean-dynamics-under-extreme-sparsity/
LOCATION:Engineering 2\, Engineering 2 1156 High Street\, Santa Cruz\, CA\, 95064
GEO:37.0009723;-122.0632371
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Engineering 2 Engineering 2 1156 High Street Santa Cruz CA 95064;X-APPLE-RADIUS=500;X-TITLE=Engineering 2 1156 High Street:geo:-122.0632371,37.0009723
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20250812T100000
DTEND;TZID=America/Los_Angeles:20250812T100000
DTSTAMP:20250925T231626Z
CREATED:20250811T070000Z
LAST-MODIFIED:20250925T231626Z
UID:10000098-1754992800-1754992800@events.ucsc.edu
SUMMARY:Mawhorter\, R. (CSE) - Certified Synthesis for Interactive Media: High Assurance Metroidvania Generation
DESCRIPTION:Program verification has been applied in many contexts (including videogames)\, but the scale and complexity of the examples that have been analyzed fall short of the ability to analyze many existing games without massive computational costs. My research focuses on automatic analysis and design of one particular game: Super Metroid\, with the goal of creating general methods for efficient analysis that address these issues. In pursuit of this goal\, I develop novel abstraction strategies that can be reapplied in other contexts. I also show that these same techniques can also be used to synthesize games\, and I develop a paradigm for understanding procedural generation problems as verification problems. This paradigm enables generators to certify their output\, and these certificates act as a powerful debugging tool. My research expands on existing techniques for applying symbolic search to large state spaces\, exploring many different ways of optimizing the state space representation\, and reporting on their relative effectiveness in real-world contexts. I also demonstrate how multiple layers of abstraction can be used to enhance existing search algorithms. Using these methods\, I show how verifying properties of software with respect to the humans that interact with it can be practically achieved. \nEvent Host: Ross Mawhorter\, PhD Candidate\, Computer Science & Engineering \nAdvsior: Adam Smith
URL:https://events.ucsc.edu/event/mawhorter-r-cse-certified-synthesis-for-interactive-media-high-assurance-metroidvania-generation/
LOCATION:Engineering 2\, Engineering 2 1156 High Street\, Santa Cruz\, CA\, 95064
GEO:37.0009723;-122.0632371
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Engineering 2 Engineering 2 1156 High Street Santa Cruz CA 95064;X-APPLE-RADIUS=500;X-TITLE=Engineering 2 1156 High Street:geo:-122.0632371,37.0009723
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20250805T100000
DTEND;TZID=America/Los_Angeles:20250805T100000
DTSTAMP:20250925T231319Z
CREATED:20250801T070000Z
LAST-MODIFIED:20250925T231319Z
UID:10000086-1754388000-1754388000@events.ucsc.edu
SUMMARY:Larsen\, B. (CMPM) - Communal Narrative Play in Perennial Games
DESCRIPTION:Online communities tell stories with the games they play. As continual updates\, recurring monetization\, and platforms for community discussions have flourished\, we have seen a rise in video games using ongoing development to tell stories\, and have a community interact with those stories and build upon them. In this dissertation\, I study this phenomenon\, which I call textit{perennial games}—storytelling experiences\, which are perpetual\, continuous\, and tell an ongoing\, communal story\, where everyone influences its future in big and small ways. I study this especially as it has grown in the years 2010-2025\, as the modern rise of the live-service game has exploded in popularity\, and are using this format to tell stories in ways both unique yet also in ways that harks back to serial fiction\, professional wrestling\, modern television series\, traditional mythology\, and more. Through a three-pronged focus I study: 1) the games as narrative experiences\, and how they facilitate narrative play through their design\, 2) the communities who play them\, how and why they play with the narrative and stay in these worlds for decades\, and 3) the development\, investigating the many joys and challenges of telling an ongoing story\, following the inevitable oscillations as developers interact with the community. Through this multifaceted approach\, I illustrate how perennial games cultivate community by inherently trading their mystery for familiarity\, creating strong social bonds through the communal experience of uncovering\, cataloging and deciphering mystery. Pushed forward by the inherent myth that these games will continue to change\, the communities around them strain against and increasing lack of mystery\, both seeking the safety of their social bonds while yearning for that which brought them there in the first place. Perennial games can be alluded to a developed garden\, requiring maintenance and care\, each year taking a subtly new shape\, molded by its inhabitants and its caretakers\, always a bit more wild than anyone can manage on their own\, and as it grows the people inside it grow ever more dependent on its continued existence\, until the promise that kept them there breaks. \nEvent Host: Bjarke Larsen\, PhD Candidate\, Computational Media \nAdvisor: Elin Carstensdottir
URL:https://events.ucsc.edu/event/larsen-b-cmpm-communal-narrative-play-in-perennial-games/
LOCATION:Engineering 2\, Engineering 2 1156 High Street\, Santa Cruz\, CA\, 95064
GEO:37.0009723;-122.0632371
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Engineering 2 Engineering 2 1156 High Street Santa Cruz CA 95064;X-APPLE-RADIUS=500;X-TITLE=Engineering 2 1156 High Street:geo:-122.0632371,37.0009723
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20250725T110000
DTEND;TZID=America/Los_Angeles:20250725T110000
DTSTAMP:20250925T231317Z
CREATED:20250709T070000Z
LAST-MODIFIED:20250925T231317Z
UID:10000065-1753441200-1753441200@events.ucsc.edu
SUMMARY:Basu\, S. (CSE) - Decomposition Techniques for Web-Scale Networks: Bridging Theory and Practice
DESCRIPTION:Decompositions of large-scale networks are central to many applications in graph mining\, network science\, and algorithm design. Over several decades\, a rich body of work has developed techniques to partition networks with various different objectives. However\, a noticeable gap persists between methods with strong theoretical guarantees\, and those that perform well in practice. Practical algorithms often lack provable guarantees\, while theoretically sound methods are rarely straightforward to implement\, and scale poorly. This thesis presents a suite of techniques that aim to close this gap\, presenting decomposition methods that are both theoretically grounded and practically efficient. \n\nThe first part of the thesis focuses on dense subgraph decomposition\, a fundamental problem with numerous real-world applications and a close relation to the problem of community detection studied in the complex networks literature. Specifically\, we focus on decompositions that produce a large number of small components; a variant that existing techniques struggle with. We propose a novel shift in perspective: rather than approximate the optimum closely using traditional optimization approaches\, we develop fast algorithms with provable lower bounds on output quality. We introduce some new objectives and metrics to achieve these\, and introduce a new theoretical framework that captures structural properties of real-world networks. We compare our perspective with the common approach taken in the community detection literature\, and demonstrate algorithms that significantly outperform prior methods across a broad range of datasets. \nThe second part of the thesis explores how network decompositions can enable sublinear-time algorithms: ones that produce approximate solutions without needing to inspect the entire input. We study two somewhat distinct problems. The first is a property testing problem in bounded-degree planar graphs\, where we show that hyperfinite decompositions allow for efficient testing of even the most complex properties. The second examines shortest-path computation in real-world networks. By observing that shortest paths often traverse a dense core\, we design the first sublinear algorithm that exploits this structure to approximate shortest paths. Our method competes with exact algorithms in speed\, while scaling to larger networks that such algorithms cannot handle. \nWe leave some open problems and discuss ongoing and future work\, and provide pointers on how our insights can be leveraged to study a larger class of problems in graph algorithms. \nEvent Host: Sabyasachi Basu\, PhD Candidate\, Computer Science & Engineering \nAdvisor: C. Seshadhri
URL:https://events.ucsc.edu/event/basu-s-cse-decomposition-techniques-for-web-scale-networks-bridging-theory-and-practice/
LOCATION:Engineering 2\, Engineering 2 1156 High Street\, Santa Cruz\, CA\, 95064
GEO:37.0009723;-122.0632371
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Engineering 2 Engineering 2 1156 High Street Santa Cruz CA 95064;X-APPLE-RADIUS=500;X-TITLE=Engineering 2 1156 High Street:geo:-122.0632371,37.0009723
END:VEVENT
END:VCALENDAR