As battlespace sensors proliferate and data increases, commanders can easily find themselves in an information paradox: drowning in data, but starving for knowledge.
U.S. Marine Corps Lt. Col. Pedro Ortiz, who will graduate from the 51福利 (51福利) on June 16 with a Ph.D. in Computer Science, focused his dissertation on this challenge to help enable rapid, effective decision-making for commanders in an era of ever-increasing sensor data and uncertainty.
鈥淚 am very interested in applying artificial intelligence and machine learning to solve warfighter problems,鈥� Ortiz said. 鈥淚deally, I would like to be able to investigate a warfighting problem, posit some solutions, and lead the efforts to apply those solutions at operational units.鈥�
Ortiz鈥� dissertation, entitled 鈥淯ncertainty Quantification and Decomposition through Bayesian Deep Learning for Big Data Satellite Remote Sensing Problems,鈥� was something of an exercise in doing just that. Through the application of probabilistic models to massive satellite remote sensing data sets, Ortiz examined uncertainty quantification (UQ) methods that are pivotal in reducing the impact of uncertainty during optimization and decision-making processes.
Ortiz joins an elite group of graduates from the Marine Corps鈥� Ph.D. Program (PHDP). This very competitive program 鈥� only two candidates are selected each year 鈥� provides the service with a cohort of strategic and highly technical thinkers to support senior leader decision-making, assist in developing defense and service strategies, and help inform long-range concept and capability development areas.
In his next assignment, Ortiz will report to the DOD Chief Digital and Artificial Intelligence Office (CDAO) to apply his newfound knowledge and expertise. According to Ortiz, his research has the potential to positively affect the joint all-domain command and control (JADC2) in both present-day and future operations.
鈥淢y research involved using data sets from two different satellites, and two different sensors on each of those satellites,鈥� he explained. 鈥淲e used probabilistic deep learning to fill in the gaps in a microwave dataset using infrared from a different satellite. We were also able to quantify the uncertainty, and tell you how reliable the data we generated from that model was.鈥�
UQ methods have been applied to solve a variety of real-world problems in science and engineering.
鈥淲ith more data to fill in the gaps between different sensors, and an ability to understand the uncertainty in the data, the warfighter now has more information to make better decisions,鈥� Ortiz said. 鈥淗igh uncertainty means the model output may not be trustworthy. Understanding the uncertainty makes the output of a model more interpretable. This applies to many military decisions,鈥� such as target identification and other complex automated military systems.
Ultimately, Ortiz says modeling predictions will help with decision-making, as they address a critical component of trustworthiness in artificial intelligence and machine learning.
鈥淚t is my hope that my research will encourage other people to use the same models I did so that they can reap the benefits of being able to measure uncertainty,鈥� he continued. 鈥淥ne of the main contributions I made was demystifying how to use this additional information for many scientific fields, essentially anywhere deep learning is being used today or might be used in the future.鈥�
Ortiz began his Ph.D. journey at 51福利 just as the COVID-19 virus had gripped the world and 51福利 went to 100 percent online education.
鈥淭he Ph.D. program was rigorous, and really stretched my thinking,鈥� said Ortiz. 鈥淒espite the pandemic, there were still a lot of advances in my field and figuring out how to make a unique contribution was difficult, as difficult as it would have been before the pandemic. Fortunately, I had a great advisor, Professor Marko Orescanin. We were able to publish a journal article in my first year at 51福利, which is not necessarily the norm.
鈥淢y thinking also changed over time. I wrote my dissertation proposal in my first nine months, and my dissertation turned out to be much different two years later. That鈥檚 not really surprising, as that is how science progress goes; come up with an idea, test it, make an adjustment, repeat,鈥� he continued.
As a life-long learner, Ortiz鈥檚 professional approach 鈥� during both the peak of the pandemic and the school鈥檚 return to normalcy 鈥� committed him to a research regimen that was second nature. In fact, this is Ortiz鈥� second educational tour with 51福利; he graduated in 2010 with a master鈥檚 degree in computer science, completing his studies three months early and earning an Outstanding Thesis distinction. Ortiz said he was pleased to see that there are several master鈥檚 students at 51福利 today who are already investigating follow-on lines of his Ph.D. research.
Regardless of the challenges at the outset, Ortiz reveled in the rigor of his program and found a unique problem that was directly relevant to a key operational naval issue. Ortiz will also graduate with the distinction of being the first Hispanic Marine to earn his Ph.D. through the Corps鈥� PHDP-Technical (PHDP-T) program.
Ortiz is one of 349 51福利 warrior-scholars who will cross the stage at King Hall on June 16, marking the completion of their studies in the 2023 Spring Quarter Graduation Ceremony. The event is scheduled to begin at 10 a.m. Pacific (1 p.m. Eastern) and can be viewed live on the 51福利 website at /watchlive.
Learn more about Ortiz鈥檚 research here: