The USRA Aviation Data Science Lab is conducting research and education to advance national capabilities for applying machine learning and other data science techniques to grand challenges in aviation systems such as Urban Air Mobility. In collaboration with NASA’s Ames Research Center, academia and industry, a focus of the lab is on developing curriculum on aviation data sciences with academia for workforce development, as well as research and development to advance national needs for improved mobility of people and goods. Advances in Artificial Intelligence and Data Science are already positively impacting air transportation of goods and people, and have the promise for increased impact as the number of air vehicles grows exponentially with unmanned aerial systems and the amount of data grows in civil aviation with connected aircraft.
This course teaches introductory and advanced methods in unsupervised learning and reasoning as well as natural language processing and its application to the aviation domain. Participants will learn to reason through a vast amount of unlabeled and unstructured data and process it for down-stream tasks such as representation learning, clustering, and anomaly/outlier detection.
Course will be taught in two phases:
Pre-Requisites
This course teaches you advanced topics in machine learning such as detailed implementation of deep neural networks with application to the aviation domain. We will be using Sherlock, FAA, BTS, and FOQA data to build case studies throughout the course around applications of interest in predictive models for air traffic management, airport surface operations, and flight safety.
Course will be taught in two phases:
Pre-Requisites
This course teaches you fundamentals of reproducible data science and analytics, probabilistic reasoning & statistical inference, and machine learning to leverage data generated within the large scope of aviation and aeronautics.
Course will be taught in four modules:
We will cover three main areas of aviation data:
We will be using ATD-2, Sherlock, and FOQA data to build the case studies.
Course is designed in two phases: (i) lecture and discussion: on the important topics in each module, and (ii) lab: with implementation of the methods learned on the real-world data using Python in Jupyter Hub. Evaluation will be based on a few individual assignments and a group project.
Pre-Requisites
You can learn basics of this topic here: http://www.cs.cmu.edu/~zkolter/course/linalg/index.html.
Many online short-courses are available to learn the basics.
USRA-NASA-Berkeley Aviation Data Science Seminar 2020.