This repository doubles as a Final Project and coursework repository for MIT OCW 18.065: Matrix Methods in Data Analysis, Signal Processing, and Machine Learning.
FinalProject contains all the source code, documentation, and results for my project.
In recent years the growing number of satellite launches and an advancement in satellite sensing capabilities has led to an increase in the amount of data collected and processed in space. In particular earth observation (EO) satellites are set to triple over the next decade. EO satellites can currently downlink 100 TB of data per day but advancements in resolution capability or new sensing modalities may increase this number significantly. These satellites are constrained by a downlink window of less than ten minutes. If the size of downlink data is too large, important data may not be transmitted to the ground in time. Image compression is an approach to solve this problem. The purpose of this project is to test the compression performance of two lossy compression techniques, Singular Value Decomposition (SVD) and Convolutional Autoencoders (Conv AE), on satellite image data.
If this seems interesting to you I recommend reading the full report to learn more about the project objectives, approach, and results.
MiniLessons contains "from-scratch" Python implementations of the Recursive Least Squares algorithm and Krylov-Arnoldi iteration. It also contains unfinished notebooks for convolution, a topic I would like to teach in the future.
PSETS includes some helper code for problem sets but does not include most of my major problem set solutions. These are in physical notebooks and will not be published on GitHub.
Contact: [email protected]
Please reach out if you would like to learn more about what I learned in this course or if you have ideas for projects in the data science/machine learning field! I would love to help if I can.