honors thesis machine learning

Navigation Menu

Search code, repositories, users, issues, pull requests..., provide feedback.

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly.

To see all available qualifiers, see our documentation .

• Notifications You must be signed in to change notification settings

This is a project for Spencer Ashford's BYU Undergraduate Honors Thesis: Applying Machine Learning Methods to Denoise HoloOcean Simulated Profiling Sonar Images

sashford/Honors-Thesis-Spencer-Ashford

Folders and files.

Repository files navigation

Honors thesis - applying machine learning to denoise holoocean simulated sonar images.

Note: Thesis was not completed as it was delaying graduation, however, this repo will remain to document my work and leave a place to pick up on my own time.

Sonar is an integral part of underwater robotics as it provides information in environments where vision is limited and many types of traditional sensors fail to give reliable results. One of the downsides of sonar is that it tends to have noise or artifacts that reduce the clarity of the data. BYU's HoloOcean underwater robotics simulator has sought to replicate this noise to give a more realistic depiction of sonar in its simulations. This thesis seeks to use a deep learning model to demonstrate a solution to removing noise from the simulated sonar data generated by HoloOcean.

The dataset consists of over 1000 image pairs generated from simulated sonar imagary using BYU's HoloOcean underwater robotics simulator. Each image pair consists of a "noisy" and "noiseless" image, corresponding to a single profiling sonar output with simulated noise either added or not respectively. These images were then converted into 512x90 numpy (.npy) files for convenient storage and usage. The dimensions 512x90 correspond to the images generated in the simulation with 512 range bins and 90 azimuth bins that the simulation used to define its resolution.

Training Script

The training script consists of a python notebook meant to work with Google Colab and make use of their hardware acceleration capabilities. It should be possible to use other IDEs and to change the hardware accelerator as needed, though this will require individual adaptation.

HonorsThesis.ipynb

This script makes use of a traditional CNN denoising autoencoder. This model runs but its effectiveness is limited by the fact that this dataset is so sparse which can be a challenge for CNNs.

TorchSParseDenoiser.ipynb

This script attempts to use the MinkowskiEngine to do sparse convolution. This proves promising for this application but I struggled to get a working implementation.

MinkowskiEngine

This is a really helpful library for doing sparse convolution and generally doing machine learning on several types of sparse data use cases such as point clouds. It does not seem fully tailored for the application this repo focuses on but seems to have potential if it can be worked through. Documentation can be found at https://nvidia.github.io/MinkowskiEngine/ and the repo can be found at https://github.com/NVIDIA/MinkowskiEngine .

• Jupyter Notebook 99.3%
• Python 0.7%

• < Previous

Home > CIPS > epo > HSP > Honors Sch Theses > 969

Honors Scholar Theses

Probabilistic machine learning for battery state of health prognostics.

Charli Zaretsky Follow

Date of Completion

Spring 5-1-2023

Thesis Advisor(s)

Chao Hu; Jason Lee

Honors Major

Mechanical Engineering

• Disciplines

The ability to understand and predict the state of health (SOH) of lithium-ion batteries is an integral component of their widespread commercial use. There are various methods through which SOH can be analyzed and predicted, and this paper discusses these different methods, and the strengths and weaknesses of each. This paper also details an analysis of lithium-ion battery SOH through two data-driven machine learning methods: XGBoost and Gaussian process regression. A comparison is made between each method’s accuracy in predicting next-cycle discharge capacity using electrochemical impedance spectroscopy (EIS) readings and battery charge and discharge rates, from a dataset given in a Nature Communications Journal article. In this application, both methods show similar results in prediction accuracy, and the use of one over the other depends heavily on individual needs, including a need for a confidence interval of the prediction, the allotted amount of time, and prior knowledge of the data.

Recommended Citation

Zaretsky, Charli, "Probabilistic Machine Learning for Battery State of Health Prognostics" (2023). Honors Scholar Theses . 969. https://digitalcommons.lib.uconn.edu/srhonors_theses/969

Since June 26, 2023

Included in

Mechanical Engineering Commons

Advanced Search

• Notify me via email or RSS
• Using and Citing
• UConn Library
• Open Access Author Fund
• Collections

Author Corner

• Submit Research
• HSP Website

Home | About | FAQ | My Account | Accessibility Statement

Privacy Copyright

A Showcase of scholarship, research, and creativity at the university of southern mississippi

• < Previous

Home > Honors College > Theses > 730

Honors Theses

Impact of artificial intelligence, machine learning, and automation in operations management: an analysis of healthcare, manufacturing, and retail sectors.

Jasmine N. Kelley

Date of Award

Degree type.

Honors College Thesis

Management and International Business

First Advisor

Ali K. Dogru, Ph.D.

Advisor Department

In this thesis, we study the impacts of artificial intelligence (AI), machine learning (ML), and automation on operations management (OM). We specifically target the industries of healthcare, retail, and manufacturing in our analysis since each sector has distinct dynamics and concerns regarding the structural changes imposed by advancements in AI. We analyze 120 high-quality peer-reviewed journal articles to capture recent research trends in healthcare, retail, and manufacturing. Using respected business analytics reports, we compare academic efforts to industry initiatives. We find that growing interest in AI research in academia is consistent with growing research and development (R&D) investment and patent acquisition on AI technologies. We conclude that the industry that is likely to be impacted the most is manufacturing due to the large amounts of machinery and technology that are used in production, followed by retail and healthcare. We show that the impacts of AI, ML, and automation will change the landscape of OM and how businesses operate and make decisions in the near future. We provide some insights for students, academic institutions, and business organizations that operate in healthcare, manufacturing, and retail sectors. We recommend possible avenues of exploration to researchers interested in AI research.

Copyright for this thesis is owned by the author. It may be freely accessed by all users. However, any reuse or reproduction not covered by the exceptions of the Fair Use or Educational Use clauses of U.S. Copyright Law or without permission of the copyright holder may be a violation of federal law. Contact the administrator if you have additional questions.

Recommended Citation

Kelley, Jasmine N., "IMPACT OF ARTIFICIAL INTELLIGENCE, MACHINE LEARNING, AND AUTOMATION IN OPERATIONS MANAGEMENT: AN ANALYSIS OF HEALTHCARE, MANUFACTURING, AND RETAIL SECTORS" (2020). Honors Theses . 730. https://aquila.usm.edu/honors_theses/730

Since September 03, 2020

Included in

Business Administration, Management, and Operations Commons

Advanced Search

• Notify me via email or RSS
• Collections

Author Corner

Home | About | FAQ | My Account | Accessibility Statement

Privacy Copyright

Home > School, College, or Department > Honors > Honors Theses > 1549

University Honors Theses

Enhancing robustness of machine learning models against adversarial attacks.

Ronak Guliani , Portland State University Follow

First Advisor

Nirupama Bulusu

Date of Award

Spring 6-2024

Document Type

Degree name.

Bachelor of Science (B.S.) in Computer Science and University Honors

Computer Science

Adversarial Attacks, Machine Learning Robustness, Adversarial Training, Model Resilience

10.15760/honors.1581

Machine learning models are integral for numerous applications, but they remain increasingly vulnerable to adversarial attacks. These attacks involve subtle manipulation of input data to deceive models, presenting a critical threat to their dependability and security. This thesis addresses the need for strengthening these models against such adversarial attacks. Prior research has primarily focused on identifying specific types of adversarial attacks on a limited range of ML algorithms. However, there is a gap in the evaluation of model resilience across algorithms and in the development of effective defense mechanisms. To bridge this gap, this work adopts a two-phase approach. First, it simulates attacks like the Basic Iterative Method (BIM), DeepFool, and Fast Gradient Sign Method (FGSM)on common ML models trained on MNIST and CIFAR-10, which are used for image processing. This thesis will then discuss defensive strategies, which reduces the sensitivity of the model to input changes to improve model resilience against attacks. The findings are aimed to benefit ML researchers to develop more secure and robust ML systems.

Persistent Identifier

https://archives.pdx.edu/ds/psu/42187

Recommended Citation

Guliani, Ronak, "Enhancing Robustness of Machine Learning Models Against Adversarial Attacks" (2024). University Honors Theses. Paper 1549. https://doi.org/10.15760/honors.1581

Since June 17, 2024

Included in

Artificial Intelligence and Robotics Commons , Theory and Algorithms Commons

Advanced Search

• Notify me via email or RSS
• Featured Collections
• All Authors
• Schools & Colleges
• Dissertations & Theses
• PDXOpen Textbooks
• Conferences
• Collections
• Disciplines
• Faculty Expert Gallery
• Submit Research
• Faculty Profiles
• Terms of Use
• Feedback Form

Home | About | My Account | Accessibility Statement | Portland State University

Privacy Copyright

• Undergraduate Honors Thesis

Investigation of Machine Learning for Jet Momentum Reconstruction in Heavy Ion Collisions Public Deposited

This thesis investigates the application of machine learning methods to the reconstruction of transverse momentum of jets in heavy ion collisions. This project largely follows from the paper  Machine-learning-based jet momentum reconstruction in heavy-ion collisions   from Haake and Loizides, in which they simulate heavy ion collisions and train three machine learning algorithms that reconstruct jet momentum .  My project investigates this methodology and the validity of these results, explores the viability of linear regression as a transparent alternative to machine learning, and considers whether machine learning is an appropriate tool for jet momentum reconstruction.  We provide an overview of useful background information including: quark- gluon plasma, jets, and machine learning.  Following is a detailed review of the methodology, covering the software used and specific implementation.  Results for three simulated jet momentum distributions are presented with further analysis of a baseline model using linear regression. Finally, we r eflect on the implications of this project and suggests avenues for future investigations.

• Lang, Jordan D.
• Nagle, James L
• Cumalat, John P
• Vernerey, Divya E.
• University of Colorado Boulder
• Machine learning
• Nuclear and particle physics
• Relativistic heavy ion physics
• Jet momentum reconstruction
• High energy nuclear experiment
• In Copyright
• English [eng]
• Creative Commons BY Attribution 4.0 International

Relationships

• Arts and Sciences Honors Program

ASU honors student explores the depths of neuroscience, machine learning

Katrina Ager is earning a Bachelor of Science in neuroscience.

Editor’s note:  This story is part of a series of profiles of  notable spring 2024 graduates .

Katrina Ager's academic journey has been fueled by a deep curiosity about the human mind. 

From understanding cognition to unraveling behavior, she has been enthralled by the complexities and mathematical nature of the brain. This passion led her to pursue a Bachelor of Science in neuroscience from ASU’s Department of Psychology, a unit within The College of Liberal Arts and Sciences, as well as certificates in applied business data analytics and computational life sciences .

During her academic tenure, Ager completed an internship at the world-famous Allen Institute for Neural Dynamics , where she contributed to the OpenScope Databook , a resource that addresses reproducibility issues and teaches computational analysis techniques to analyze publicly available data. This experience amplified her passion, particularly at the intersection of neuroscience and technologies like machine learning.

"There is so much about the brain that remains a mystery, and it may not be understood in my lifetime, but it is something that will captivate me for the rest of my life," Ager said. “I have always been fascinated by the human condition: what it means to think, feel, love, interact, contemplate and experience life in deep ways.”

A member of Barrett, The Honors College , Ager excelled, maintaining a 4.0 GPA and earning a place on the Dean's List each semester. Financial assistance from programs like ASU’s President Barack Obama Scholars Program and the New American University President’s scholarship supported her academic journey.

Read more about Ager’s ASU experience and plans moving forward.

Question: What’s the best piece of advice you’d give to those still in school?

Answer: Never judge anyone before you get to know them and try to be curious about other people's experiences and perspectives. I have met some of my best friends and learned so much from people I might never have interacted with if I had not actively tried to get to know my peers. I believe you can learn something from everyone you meet, and it's important to be open to new perspectives and friendships during a time when everyone is learning about themselves and life.

Question: Was there a specific professor or mentor that significantly impacted your time at ASU?

Answer: Professor Samuel McClure has, without a doubt, been the person who has taught me the most important lessons and supported me throughout my academic career. If you talk to anyone who has taken his class, they would agree that Dr. McClure is a person who truly cares about his students and provides the support they need to excel academically and in life. Dr. McClure has continuously inspired me, shown me the value of working hard and staying curious, and helped me achieve my goals inside and outside of the classroom. I think this campus is a better place with people like him, and I am extremely grateful to have had his mentorship over the past four years.

Question: Were there any specific experiential learning opportunities that significantly influenced your academic and personal growth during your time at ASU?

Answer: Working as a lab manager for Dr. Gene Brewer ’s Memory and Attention Control Lab at ASU has significantly impacted my academic journey. Exploring different research labs as an undergraduate was essential to figuring out what interested me. I am fortunate to be surrounded by a supportive and caring team at the MAC Lab, and I genuinely look forward to going to work and interacting with my lab mates. Being a lab manager has contributed to my understanding of how research is conducted, taught me how to ask powerful questions to gain insights and how to work effectively with a team. 

Question: What did your honors thesis explore?

Answer: For my honors thesis, I worked under the mentorship of Dr. McClure to understand how a new machine learning model lets us measure neurotransmitter concentrations in the human brain. The model was developed by Dr. Read Montague at Virginia Tech, and his team has been incredibly supportive in allowing me to use their model and data for my thesis. This research is groundbreaking, as it can track sub-second changes in concentrations of neurotransmitters such as dopamine, norepinephrine and serotonin.

Question: What is something you learned while at ASU — in the classroom or otherwise — that surprised you or changed your perspective?

Answer: One thing that has changed my perspective has been recognizing the importance of interacting with my peers, truly listening when they speak and approaching everything with a curious mind. During my sophomore year, I took a special topics class on dopamine with Professor Samuel McClure and Assistant Research Professor Kimberlee D’Ardenne that revealed the importance of community for me. It was one of the first in-person classes I had after being on Zoom for a year-and-a-half due to COVID, and everyone was extremely eager to connect face-to-face. Aside from being very informative and interesting academically, the happiness, support and genuine interest in connection among people in the class were unmatched. 

That excitement bred many interesting conversations and led to a deeper understanding about the topics being discussed. I absorbed so much knowledge from listening to my friends in the class. It has made me put effort into approaching other areas of my life with the same curiosity. That class has been instrumental in my education because it made me realize that no matter what I end up doing with my degree, as long as I am surrounded by supportive, curious and passionate people, I will be just fine.

Question: Can you share more about your plans after graduation?  

Answer: I am excited to have recently accepted a position at the Dynamical Inference Lab at the Institute of Computational Biology of Helmholtz Munich. Under the mentorship of Steffen Schneider, I will be gaining experience in software development, computational neuroscience and machine learning. I also plan to simultaneously pursue a master’s degree while I am working abroad. I am eager to move to Germany in the fall and will be staying for one to two years. In the future, I hope to transition into industry and work for a company — or start one myself — that is grounded in a humanitarian mission while being on the cutting edge of science and technology.

More Sun Devil community

Real Time Analysis Center enhances protection of community, resources

It’s a nightmare scenario: A report of an active shooter at a building on a bustling university campus.But instead of panic, the reaction is a swift, mediated assessment of real-time footage within…

From the high wire to the pool

Arizona State University swimming and diving coach Herbie Behm didn’t truly know how well-known Ilya Kharun’s family is until he checked the team into a Seattle-area hotel for the Pac-12…

ASU Alumni Association ranks as No. 1 largest Phoenix-area networking association

With more than 640,000 members, the ASU Alumni Association was recently ranked the No. 1 largest Phoenix-area networking association by the Phoenix Business Journal for the 11th consecutive year.…

File(s) under embargo

until file(s) become available

PHYSICS-GUIDED MACHINE LEARNING APPLICATIONS FOR AIR TRAFFIC CONTROL

The Air Traffic Management (ATM) system encompasses complex and safety-critical operations which are mainly managed by Air Traffic Controllers (ATCs) and pilots to ensure safety and efficiency. This air traffic operation becomes more complex and challenging as demands continue to increase. Indeed, the demand for air transport is expected to increase by an average of 4.3% annually over the next 20 years, and the projected number of flights is expected to reach around 90 million by 2040 [1]. This continuous growth of demands can lead to an excessive workload for both ATCs and pilots, thereby resulting in the degradation of the ATM system. To effectively respond to this problem, a lot of effort has been put into developing decision support tools. This dissertation explores and focuses on the development of algorithms for decision support tools in air traffic control, emphasizing specific desirable properties essential for tasks such as tracking the position of aircraft and monitoring air traffic. The primary focus of this dissertation is to combine a data-driven model and a physics-based model systematically, thereby addressing the limitations of previous works in trajectory prediction and anomaly detection. Through a literature review, important properties, including real-time applicability, interpretability, and feasibility, are identified and pursued for practical applications. These properties are integrated into the proposed algorithms which combine data-driven and physics-based models to address dynamic air traffic scenarios effectively. To meet the requirement of real-time applicability, the algorithms are designed to be computationally efficient and adaptable to continuously changing conditions, ensuring timely provision of immediate information and near-instantaneous responses to assist ATCs. Subsequently, interpretability allows controllers to understand the reasoning behind the algorithm’s predictions. This is facilitated by the use of attention mechanisms and explicit physics-based guidance, making the predictions more intuitive and understandable. In addition, anomaly detection algorithms provide human-readable decision boundaries for flight states for a clear understanding. Lastly, feasibility ensures that the algorithms generate realistic aircraft trajectory predictions based on current flight states and air traffic conditions. This is achieved by physics-guided machine learning which leverages both data-driven and physics-based approaches, accounting for the aircraft dynamics and uncertainties. Moreover, practical and operational considerations are integrated into algorithms for real-world applications. This includes developing anomaly detection models that are adaptable to dynamic trajectory patterns to address the complexities of flexible area navigation airspace. Additionally, to reduce the workload of ATCs, providing immediate advisories for anomaly resolution and arrival sequencing is targeted by learning from historical data. By considering these properties with practical considerations, the dissertation presents a suite of algorithms that can effectively support human operators for air traffic control.

Degree Type

• Doctor of Philosophy
• Aeronautics and Astronautics

Campus location

• West Lafayette

Advisor/Supervisor/Committee Chair

Additional committee member 2, additional committee member 3, additional committee member 4, usage metrics.

• Aerospace engineering not elsewhere classified