Date of Award:
8-2023
Document Type:
Thesis
Degree Name:
Master of Science (MS)
Department:
Computer Science
Committee Chair(s)
Soukaina Filali Bourbrahimi
Committee
Soukaina Filali Bourbrahimi
Committee
Mario Harper
Committee
Mike Taylor
Abstract
Neural networks are adept at finding patterns that are too long and too small for humans to find in data. Usually, this power is used to generate predictions with greater accuracy than most alternative models. However, we can also use this power to understand more about the data we train these networks on. We do this by changing the data that the networks train on and the data they are tested on. This allows us to both control the maximum length of a pattern and to compare data between different groups, in our case, different solar cycles. This thesis is our attempt to understand solar wind data better. We do this by proposing a physics based framework and comparing the results of different inputs and outputs through different networks. These results show three major things: 1, that training networks using the physical law of Ohm's law for an ideal plasma can improve network performance predictions; 2, that the specific characteristics of different solar cycles make them more suitable for training or testing; and 3, that while physics guided loss functions can be helpful in certain situations, they are no silver bullet to improved predictions.
Checksum
415c0628b39bec8e1ecf216619cb6aab
Recommended Citation
Johnson, Robert M., "Physics-Guided Deep Learning for Solar Wind Modeling at L1 Point" (2023). All Graduate Theses and Dissertations, Spring 1920 to Summer 2023. 8871.
https://digitalcommons.usu.edu/etd/8871
Included in
Copyright for this work is retained by the student. If you have any questions regarding the inclusion of this work in the Digital Commons, please email us at .