Hierarchical Federated Learning On Healthcare Data: An Application To Parkinson's Disease,
2023
Air Force Institute of Technology
Hierarchical Federated Learning On Healthcare Data: An Application To Parkinson's Disease, Brandon J. Harvill
Theses and Dissertations
Federated learning (FL) is a budding machine learning (ML) technique that seeks to keep sensitive data private, while overcoming the difficulties of Big Data. Specifically, FL trains machine learning models over a distributed network of devices, while keeping the data local to each device. We apply FL to a Parkinson’s Disease (PD) telemonitoring dataset where physiological data is gathered from various modalities to determine the PD severity level in patients. We seek to optimally combine the information across multiple modalities to assess the accuracy of our FL approach, and compare to traditional ”centralized” statistical and deep learning models.
Lunar Search And Rescue: The Next Step For Human Spaceflight Recovery,
2023
Air Force Institute of Technology
Lunar Search And Rescue: The Next Step For Human Spaceflight Recovery, Benjamin J. Johnis
Theses and Dissertations
Humans return to the Moon in 2024 under the NASA Artemis mission. Commercial space explorations are rapidly expanding. The sudden increase of human spaceflight calls for increased emphasis on space rescue. These revelations drove an AFRL/AFIT capability gap assessment to provide specific proposals for a modern space rescue program guided by Transaction Cost Economics Theory. The mission analysis arms USSPACECOM and NASA to coordinate an interagency approach to Lunar Search and Rescue, positioning America to compete in the new space race. The strategic consequences of failing to act could fundamentally jeopardize American leadership in space.
Automated Registration Of Titanium Metal Imaging Of Aircraft Components Using Deep Learning Techniques,
2023
Air Force Institute of Technology
Automated Registration Of Titanium Metal Imaging Of Aircraft Components Using Deep Learning Techniques, Nathan A. Johnston
Theses and Dissertations
Studies have shown a connection between early catastrophic engine failures with microtexture regions (MTRs) of a specific size and orientation on the titanium metal engine components. The MTRs can be identified through the use of Electron Backscatter Diffraction (EBSD) however doing so is costly and requires destruction of the metal component being tested. A new methodology of characterizing MTRs is needed to properly evaluate the reliability of engine components on live aircraft. The Air Force Research Lab Materials Directorate (AFRL/RX) proposed a solution of supplementing EBSD with two non-destructive modalities, Eddy Current Testing (ECT) and Scanning Acoustic Microscopy (SAM). Doing …
Pattern Of Life Modeling On Maritime Vessels With Bayesian Inferencing,
2023
Air Force Institute of Technology
Pattern Of Life Modeling On Maritime Vessels With Bayesian Inferencing, Miles A. Kelly
Theses and Dissertations
Despite decades of research, maritime traffic models’ limited predictive power continues to constrain their operational utility. We build on previous pattern of life modeling algorithms and contribute a Bayesian inferencing model for anomaly detection. Our probabilistic approach provides decision makers the capability to tailor the belief threshold for identifying anomalies and to enact a measured response based on the degree of abnormality. We perform a case study to evaluate the results, verifying that our Bayesian inferencing method accurately refines its probability when given the location and time of a ship of interest, and serving as a proof of concept for …
Uncertainty Quantification In Federated Learning For Persistent Post-Traumatic Headache,
2023
Air Force Institute of Technology
Uncertainty Quantification In Federated Learning For Persistent Post-Traumatic Headache, Byungmoo Brian Kim
Theses and Dissertations
A post-traumatic headache (PTH), resulting from a mild traumatic brain injury (mTBI), potentially develops into persistent post-traumatic headache (PPTH). Although no known cure for PPTH exists, research has shown that receiving treatment at earlier stages of PTH lowers the risk of patients developing PPTH. Previous studies have shown machine learning (ML) models capable of predicting a patient’s PTH progression, but none have considered the issue of protecting patient privacy. Due to patient privacy, ML models only have access to data within the institution. Federated learning (FL) harnesses data from separate institutions without sacrificing patient privacy as institutions can run ML …
Designing A Counter-Iads Drone Swarm: Using Evolution To Evaluate Combat Assumptions Underpinning Drone Swarm Target Assignment,
2023
Air Force Institute of Technology
Designing A Counter-Iads Drone Swarm: Using Evolution To Evaluate Combat Assumptions Underpinning Drone Swarm Target Assignment, Olin H. Kennedy
Theses and Dissertations
The original research goal was to combine the best techniques in the drone swarm literature and model a functional combat drone swarm that conducts a Suppression of Enemy Air Defense (SEAD) mission. However, the body of literature regarding Drone Swarm Target Assignment (DSTA) does not model enemy counteraction and assumes that the drones’ targets are compliant against destruction. Therefore, a model of enemy counteraction against drone swarms is developed, and Novel DSTA (NDSTA) is proposed to respond to the weaknesses of the current DSTA. Both methods of target assignment are combined with a tunable trajectory generation model, and the performance …
Bayesian Recurrent Neural Networks For Real Time Object Detection,
2023
Air Force Institute of Technology
Bayesian Recurrent Neural Networks For Real Time Object Detection, Stephen Z. Kimatian
Theses and Dissertations
Neural networks have become increasingly popular in real time object detection algorithms. A major concern with these algorithms is their ability to quantify their own uncertainty, leading to many high profile failures. This research proposes three novel real time detection algorithms. The first of leveraging Bayesian convolutional neural layers producing a predictive distribution, the second leveraging predictions from previous frames, and the third model combining these two techniques together. These augmentations seek to mitigate the calibration problem of modern detection algorithms. These three models are compared to the state of the art YOLO architecture; with the strongest contending model achieving …
Advancing Autonomous Swarm Behavior In A Simulated Anti-Access Area Denial (A2ad) Environment,
2023
Air Force Institute of Technology
Advancing Autonomous Swarm Behavior In A Simulated Anti-Access Area Denial (A2ad) Environment, Chad P. Macwilkinson
Theses and Dissertations
Advancements in modern IADS have bolstered A2AD environments and subsequently degraded the advantages that the Air Force once held, prompting a call to reform the nature of warfare in order to challenge these threats. A solution is weapon swarming technology, which has the ability to overwhelm IADS by engagement of a large numbers of low-cost, but lethal air assets that have autonomous functionalities. This research proposes the application of a four dimensional framework for autonomy to a swarm of cruise missiles. A virtual A2AD environment of two opposing forces is constructed using the AFSIM, wherein a manned bomber seeks to …
Air Force Cadet To Career Field Matching Problem,
2023
Air Force Institute of Technology
Air Force Cadet To Career Field Matching Problem, Ian P. Macdonald
Theses and Dissertations
This research examines the Cadet to Air Force Specialty Code (AFSC) Matching Problem (CAMP). Currently, the matching problem occurs annually at the Air Force Personnel Center (AFPC) using an integer program and value focused thinking approach. This paper presents a novel method to match cadets with AFSCs using a generalized structure of the Hospitals Residents problem with special emphasis on lower quotas. This paper also examines the United States Army Matching problem and compares it to the techniques and constraints applied to solve the CAMP. The research culminates in the presentation of three algorithms created to solve the CAMP and …
Simulating Autonomous Drone Swarm Behaviors In An Anti-Access Area Denial (A2ad) Environment,
2023
Air Force Institute of Technology
Simulating Autonomous Drone Swarm Behaviors In An Anti-Access Area Denial (A2ad) Environment, Alexander L. Martinez
Theses and Dissertations
Army senior military leaders are invested in acquiring modernized aerial platforms and equipment to augment the U.S. Army’s ability to overcome A2AD threats imposed by modern IADS. A prominent element of this modernization effort is the employment of autonomous drones to defeat IADS threats while minimizing risk to Army Soldiers. This research utilizes a framework for classifying the levels of autonomous capability along three dimensions: the ability to act alone, the ability to cooperate, and the ability to adapt. A virtual combat model, created using the AFSIM, simulates the engagement between an enemy IADS and a friendly formation comprised of …
An Approximate Dynamic Programming Approach For Solving An Air Combat Maneuvering Problem With Directed Energy Weapons,
2023
Air Force Institute of Technology
An Approximate Dynamic Programming Approach For Solving An Air Combat Maneuvering Problem With Directed Energy Weapons, Elisha A. Palm
Theses and Dissertations
Performing within visual range (WVR) air combat involves the execution of complex air maneuvers and rapid sequential decision making. The complexity of these decisions can increase even further when including additional weapon capabilities. The advancement of unmanned autonomous vehicle technology and weapon capabilities can help combat the hindrance that comes with human limitations. Autonomous unmanned combat aerial vehicles (AUCAVs) and the implementation of advanced weapon capabilities such as Directed Energy Weapons (DEWs) can prove to be vital in a WVR air combat context. This derives the question – Can AUCAV’s possess the proper artificial intelligence and weapon capabilities to attain …
The U.S. Army Officer-To-Unit Assignment Problem,
2023
Air Force Institute of Technology
The U.S. Army Officer-To-Unit Assignment Problem, Andrea L. Phillips
Theses and Dissertations
Every two to three years, U.S. Army officers must change duty stations, which entails a selection process based on preferences. Currently, officers are assigned to units using a stable-marriage algorithm. Two impracticalities occur within this process. First, officers are required to submit strictly ranked preferences, not allowing indifference among units. Second, the stable-marriage algorithm does not give flexibility to alternative priorities. This research focuses on two modifications to the current model. First, a mixed integer program is created that allows the user, U.S. Army Human Resources Command, to consider other priorities: unit preferences and maximum officer disappointment. Second, generated data …
Simulation And Analysis Of Dynamic Threat Avoidance Routing In An Anti-Access Area Denial (A2ad) Environment,
2023
Air Force Institute of Technology
Simulation And Analysis Of Dynamic Threat Avoidance Routing In An Anti-Access Area Denial (A2ad) Environment, Dante C. Reid
Theses and Dissertations
This research modeled and analyzed the effectiveness of different routing algorithms for penetration assets in an A2AD environment. AFSIM was used with different configurations of SAMs locations and numbers to compare the performance of AFSIM’s internal zone and shrink algorithm routers with a Dijkstra algorithm router. Route performance was analyzed through computational and operational metrics, including computational complexity, run-time, mission survivability, and simulation duration. This research also analyzed the impact of the penetration asset’s ingress altitude on those factors. Additionally, an excursion was conducted to analyze the Dijkstra algorithm router’s grid density holding altitude constant to understand its impact on …
A Transformer Based Architecture For Indonesian Sentiment Analysis - Exploring Indobert Variations, Training Size, And Self-Supervised Model Training,
2023
Air Force Institute of Technology
A Transformer Based Architecture For Indonesian Sentiment Analysis - Exploring Indobert Variations, Training Size, And Self-Supervised Model Training, Connor F. Shaw
Theses and Dissertations
There is strong motivation in both civilian and military circles to understand the attitudes, motivations, feelings, and emotions of a population of interest. Social media is a rich source of self-disclosed information by individuals from all walks of life about virtually every domain of the human experience, but the vast quantity of data is impossible to effectively analyze without advanced natural language processing algorithms. This research creates a transfer learning based emotion classification model for Indonesian language Twitter data. Transfer learning consists of two steps: pre-training and fine tuning. Three variations of Indonesian Bidirectional Encoder Representations from Transformers (IndoBERT) are …
Analysis And Optimization Of Contract Data Schema,
2023
Air Force Institute of Technology
Analysis And Optimization Of Contract Data Schema, Franklin Sun
Theses and Dissertations
agement, development, and growth of U.S Air Force assets demand extensive organizational communication and structuring. These interactions yield substantial amounts of contracting and administrative information. Over 4 million such contracts as a means towards obtaining valuable insights on Department of Defense resource usage. This set of contracting data is largely not optimized for backend service in an analytics environment. To this end, the following research evaluates the efficiency and performance of various data structuring methods. Evaluated designs include a baseline unstructured schema, a Data Mart schema, and a snowflake schema. Overall design success metrics include ease of use by end …
Probability Of Agreement As A Simulation Validation Methodology,
2023
Air Force Institute of Technology
Probability Of Agreement As A Simulation Validation Methodology, Matthew C. Ledwith
Theses and Dissertations
Determining whether a simulation model is operationally valid requires the rigorous assessment of agreement between observed functional responses of the simulation model and the corresponding real world system or process of interest. This research seeks to extend and formulate the probability of agreement approach to the operational validation of simulation models. The first paper provides a methodological approach and an initial demonstration which leverages bootstrapping to overcome situations where one’s ability to collect real-world data is limited. The second paper extends the probability of agreement approach to account for second-order heteroscedastic variability structures and establishes a weighted probability of agreement …
Inducing Sparsity Within High-Dimensional Remote Sensing Modalities For Lightning Prediction,
2023
Air Force Institute of Technology
Inducing Sparsity Within High-Dimensional Remote Sensing Modalities For Lightning Prediction, Grace E. Metzgar
Theses and Dissertations
The uncertainty of lightning constantly threatens many weather-sensitive fields where the slightest presence of lightning can endanger valuable personnel and assets. The consequences of delaying operations have incited the research of methods that can accurately predict the location of future lightning strikes from the current weather conditions. High-dimensional remote sensing modalities contain information capable of detecting significant patterns and intensities within storms that could indicate the presence of lightning. This thesis induces sparsity into convolutional neural networks (CNNs) and remote sensing modalities through a combination of regularization and tensor decomposition techniques to call attention to sparse features that are most …
Critical Infrastructure System Resiliency Modeling Using Multi-Layer Network Optimization,
2023
Air Force Institute of Technology
Critical Infrastructure System Resiliency Modeling Using Multi-Layer Network Optimization, Spencer R. Figge
Theses and Dissertations
ccurately modeling the interdependent operation of critical infrastructure systems is an effective and efficient way of proactively evaluating system vulnerabilities and resiliency. Infrastructure systems are designed to transport essential commodities from where they are produced to where they are consumed and network flow-based models are one of the most effective ways to simulate and quantify infrastructure performance. The literature is populated with proposed models that must balance accuracy of interdependent operations, capability to include real-world considerations, and computational cost. This research proposes an alternative network-flow based model called the Critical Infrastructure System Resiliency Model (CISRM) that focuses on modeling a …
Optimal Control Of Precision Airdrop Trajectories Using Direct Collocation And Analytical Methods,
2023
Air Force Institute of Technology
Optimal Control Of Precision Airdrop Trajectories Using Direct Collocation And Analytical Methods, Edward J. Maxwell
Theses and Dissertations
The work herein investigates the preliminary designs of an optimal navigation controller for a scalable cylindrical airdrop system controlled with grid fins in planar motion. Precision airdrop capabilities are desired for a range of military and humanitarian missions. Fielded airdrop systems have not met desired performance objectives, particularly regarding accuracy. Direct collocation and analytical methods were utilized to solve the optimal control problem for the grid fin controlled precision airdrop system examined in this work. The optimal control problem was comprised of two phases: controlled descent and parachute descent. Minimum and maximum ranges for the system under varying wind fields …
Rate Effects In Aircraft Learning Curves,
2023
Air Force Institute of Technology
Rate Effects In Aircraft Learning Curves, Caleb J. Ahern
Theses and Dissertations
Across the Department of Defense (DoD), a wide variety of analytical tools are employed by cost analysts to estimate weapon system costs. One of the techniques widely employed by practitioners is the learning curve (LC). Although learning curves have been widely studied, using rate-adjustments or production rate effects (PRE), their usage have only intermittently been evaluated in place of using the traditional learning curve. Previous studies analyzing production rate found mixed results. This research aims to examine aircraft production data to determine if production rate model is preferrable in United States Air Force programs. Additionally, this PRE research seeks to …
