71446 research outputs found
Sort by
Optimizing the Infidelity, Sensitivity, and Complexity of Feature Importance Explanations for Machine Learning Models
No full textInterpretable machine learning aims to bridge the gap between complex model predictions and human understanding. Given the open-ended nature of the field, there is an abundance of different methods for achieving interpretability and metrics for evaluating the quality of explanations. In this thesis, we survey the existing work in the field and focus on three main types of metrics, which we refer to as \textit{infidelity}, \textit{sensitivity}, and \textit{complexity}. We explore a novel framework for interpretability that balances these three objectives using a metric of explanation quality that incorporates all three objectives, ensuring that explanations accurately capture the model's behavior, remain stable across similar inputs, and avoid being needlessly hard to interpret. Specifically, we consider explanations that attempt to generate an \textit{importance score} for each feature of the input. We calculate the optimal explanation by minimizing the metric according to an algorithm we introduce based on coordinate descent and the Adam optimizer. We implement this algorithm and evaluate our approach on a neural network trained on the MNIST dataset.Computer Scienc
Seiberg–Witten theory and reality
No full textThis thesis develops monopole Floer-theoretic invariants for webs and foams using real Seiberg–Witten theory with orbifold singularities. The approach is based on Kronheimer and Mrowka's framework of monopole Floer homology; a special case of this construction is the real monopole Floer homology for links defined by the author.Mathematic
Reclaiming Alexandria: A Case for Historic Futures
No full textOver the past centuries, scenarios of menace resulting in unchecked urban pressures have degraded the historic core of Alexandria into vertical slums and abandoned warehouses, compromising the integrity of the original Alexandrian masterplan. The complexity of this deterioration involves numerous interconnected factors, including inadequate housing, environmental degradation, mobility challenges, resource scarcity, and the lack of policy frameworks governing the dynamics among various stakeholders. Consequently, Alexandria’s historic core remains in a state of decline, awaiting a transformative vision capable of revitalizing its latent potential. This thesis, thus, proposes a reinterpretation of Alexandria’s original masterplan as a theoretical, site-specific framework translated into a tangible urban design intervention. It explores how the conservation of Alexandria’s archeological urbanism can serve as a catalyst for urban renewal and provide an example of a future-oriented pathway deeply anchored in the city’s historic identity.Department of Urban Planning and Desig
Applying Kantian Ethics to a Westernized Social Credit System
No full textDue to the popularity of new gamified systems such as summer reading programs,
language learning applications, positive behavior intervention systems, loyalty rewards
programs and more, humanity is being primed to become homo ludens, or man of play.
Due to this, inevitably, a social credit system, similar to the one that is currently
implemented in China, will become a reality in America. This thesis first determines the
elements of a social credit system that would be feasibly implemented in the United
States. Kantian ethics is used to determine the moral worth of this hypothetical system
due to its emphasis on human dignity, autonomy, community, and privacy. Every element
of a potential Westernized social credit system is assessed as either morally permissible
or impermissible under Immanuel Kant’s deontological framework. It applies his
Universal Law Test, Formula of Humanity, and the Kingdom of Ends. The thesis
concludes that if a social credit system is completely voluntary, is designed to reward
following perfect and imperfect duties, aligns incentives with intrinsic moral motivations,
does not allow people to buy points, and does not include a leaderboard, it is morally
permissible, but not a moral obligation.Extension Studie
More Than Meets the Eye: Is It Feasible To Uncover Intent From Saccade Sequences?
No full textIn 1967, Alfred Yarbus demonstrated that human eye movements are influenced by cognitive tasks, suggesting they can reveal an observer’s intent. While subsequent research produced mixed results regarding this claim, advancements in machine learning provide new ways to explore the predictability of intent from eye movement data. This thesis revisits this topic, focusing on whether saccade sequences can predict intent beyond random chance and what patterns models prioritize in these predictions. Our findings confirm that intent can be reliably predicted above chance when models are trained on targeted data, such as distinguishing between tasks where a target is present (TP) or absent (TA). Performance improvements were noted when the models specialized in either condition, although cross-scope generalization remained challenging. Temporal dependencies emerged as crucial, demonstrated by a significant accuracy drop when saccade sequences were reversed, highlighting the role of sequence order. Additionally, spatial occlusion experiments showed that specific image regions contribute differently to prediction confidence, with context-based models leveraging broader scenes and detail-based models focusing on identifiable objects. These results support Yarbus' hypothesis of purpose-driven eye movements, underscoring the intricate interplay of spatial and temporal elements in intent prediction.Computer Scienc
Beyond the Triangle: Leading District MTSS Improvement at the Speed of Trust
No full textThis Capstone explores the persistent challenge of implementing coherent, equitable Multi-Tiered Systems of Support (MTSS) in public school districts, focusing on a strategic improvement initiative in Somerville Public Schools (SPS). Despite federal mandates and widespread adoption of MTSS as a best practice, many districts struggle to move beyond symbolic compliance. As the Superintendent Fellow and utilizing a trust-centered change management approach inspired by Frances Frei and Anne Morriss, my project aimed to improve coherence in MTSS by recoupling school-based intervention blocks (X-Block), aligning data meeting protocols, and codifying district-wide expectations in a revised District Curriculum Accommodation Plan.
Through relationship-building, targeted data collection, and compelling communication, the work engaged educators throughout the district, catalyzed early shifts in practice, and laid the groundwork for long-term change. While constraints limited full-scale implementation, the project demonstrates how district leaders can build trust, generate urgency, and activate stakeholders to drive adaptive change while leveraging technical solutions. This work offers practical insights for education leaders seeking to move MTSS from theory to impactful practice.Educatio
From the Bench to the Clinic: Advancing Extracellular Matrix Material for Ligament Repair
No full textThe field of orthopedic biomaterials has advanced significantly in recent years, particularly in the development of alternatives to traditional surgical approaches for ligament and tendon injuries. Our team is currently in the process of translating an extracellular matrix-based biomaterial from the bench into the clinic for first-in-human clinical trials. The two projects presented in this thesis represent a portion of the translational process in preparation for an Investigational Device Exemption (IDE) application to the Food and Drug Administration (FDA). As part of this application, we have conducted in vitro studies optimizing a variety of parameters such as composition, concentration and sterilization. In parallel, we have conducted three large animal studies to validate the efficacy and safety of our ECM-based biomaterial. The first project in this thesis investigates the composition of our material and whether the use of whole blood versus platelet rich plasma (PRP) mixed with our ECM material, stimulates ACL fibroblast growth more. The second project investigates post-sterilization treatments to mitigate potential cytotoxic effects from residual sterilants. Both projects share a common goal: to enhance the clinical translation of biologically inspired therapies that promote tissue healing and regeneration. These efforts address critical limitations in current surgical techniques and biomaterial processing methods.Graduate Educatio
Improving Microestimates of Poverty from Satellite Images
No full textAccurately mapping the geographic distribution of poverty is pivotal for advancing development, yet this effort is often hampered by sparse, unreliable, and non-granular data. Using publicly available satellite images for nearly 20,000 villages in Africa, this paper demonstrates how self-supervised pre-training can enhance the accuracy and scalability of microestimates of poverty, as measured by the asset wealth index (AWI). This method outperforms a fully supervised machine learning approach by extracting more predictive features from the images, explaining approximately 72\% of the survey-measured variation in AWI and surpassing the current state-of-the-art by about 3 percentage points. By offering a more accurate and scalable solution for poverty estimation, this research provides valuable insights for informed policymaking and targeted poverty alleviation.Applied Mathematic
Fuel Burn Tradeoffs in Safety-Optimized Trajectories for Uncrewed Aircraft Systems
No full textThe integration of Uncrewed Aircraft Systems (UAS) into the National Airspace System presents significant operational challenges, particularly in balancing safety and fuel efficiency. Ensuring a stable Command and Control (C2) link is critical for safe operations, as Loss of C2 Link (LC2L) events can compromise situational awareness and lead to mission failure. However, minimizing LC2L risk often requires trajectory deviations, increasing fuel burn and emissions. This thesis employs simulation-based optimization methods to quantify the fuel burn penalties associated with safety-driven routing. Using a dataset of regional cargo UAS flights departing from the Dallas-Fort Worth Metroplex, three trajectory optimization strategies are evaluated: fuel-optimal, outage-minimizing, and risk-minimizing trajectories. The results indicate that prioritizing safety results in increased fuel consumption, but the magnitude of this penalty depends on factors such as transmitting power levels and ground station (GS) placement. These findings underscore the need for enhanced C2 link infrastructure, strategic placement of GS networks, and regulatory policies that balance safety and operational efficiency. This research provides actionable insights for UAS operators and policymakers, ensuring that safety-driven routing does not impose excessive fuel costs, ultimately supporting the sustainable deployment of UAS in commercial aviation.Applied Mathematic
Building On-Demand Customizable Cardiovascular Implants
No full textCardiovascular disease and injury frequently necessitate structural replacement of heart valves and blood vessels. As current clinical options often require repeat interventions, regenerative implants are of particular interest as they may enable lifelong solutions. To successfully regenerate cardiovascular components, implanted scaffolds should interact with the native environment across multiple length scales, ranging from the nanoscale of the cellular microenvironment to the centimeter scale relevant to valvular and vascular function. Current manufacturing platforms struggle to fabricate implants with concurrent control across these hierarchical length scales. This work presents Focused Rotary Jet Spinning (FRJS) as a rapid cell-free fabrication strategy capable of producing customizable cardiovascular implants comprised of nano- and microscale structural cues in just minutes. The nanofiber components of the textile scaffolds recapitulate fibrous extracellular matrix, interacting with cells at familiar length scales. When seeding representative cardiovascular cell types onto scaffolds with different alignments, the self-organizing behavior of the different cells displayed the ability to form anisotropic tissue configurations dependent on the degree of scaffold alignment. Extending these textile scaffolds to cardiovascular implants, vascular grafts were fabricated with control and customizability across nanometer to centimeter scales. When the acellular scaffolds were implanted for four weeks in rat femoral vasculature, the vascular grafts maintained patency, tissue perfusion, and initial cellular infiltration. With FRJS produced implants demonstrating the ability to convey blood flow, their ability to control these flows was examined. Pulmonary valve replacements were formed with rapid fabrication of customized scaffolds. Valve leaflet geometry parameters were adjusted, producing a valve capable of sufficiently controlling hemodynamic flows in vitro. The implant’s functionality was confirmed in a large animal model after a minimally invasive transcatheter delivery. Both valves and vascular grafts were produced in just minutes with manufacturing control across a wide range of necessary length scales. FRJS as a cell-free fabrication method provides speed and control advantages that can enable on-demand and customizable regenerative implants and lifelong solutions for patients.Engineering and Applied Sciences - Engineering Science