BYU

Abstract by Stephanie Herron

Personal Infomation


Presenter's Name

Stephanie Herron

Degree Level

Undergraduate

Abstract Infomation


Department

Physics and Astronomy

Faculty Advisor

Tracianne Neilsen

Title

Dictionary Learning for Sound Speed Profiles in Ocean Acoustics

Abstract

The speed of sound in the ocean is dependent on many variables, including temperature, density, salinity, and seasonable variables. Models of the variability of sound speed as a function of depth are called sound speed profiles (SSP). Modeling sound speed profiles is traditionally done with empirical orthogonal functions (EOFs), which are computed using principal component analysis. However, due to the requirement that the components be orthogonal, these functions fail to model small-scale fluctuations which may arise from internal waves or other disturbances in the ocean. An improvement is using Dictionary Learning, a type of unsupervised machine learning algorithm that computes a collection of basis vectors, called atoms, that are not necessarily orthogonal. These atoms can capture small fluctuations and can then be used to represent and generate realistic sound speed profiles in the ocean.