Skip to content

Jiajun-Bao/Projects

Repository files navigation

Data-Analysis-and-Machine-Learning

Note: Projects 1-5 were completed during my undergraduate studies at the University of Washington, and Project 6 was completed during my master’s studies at the University of Chicago.

Project 1: Submarine Detection Using Discrete Fourier Transform and Noise Filtering Techniques

This project solves a submarine radar detection problem, using noisy acoustic data obtained over a 24-hour period in half-hour increments to determine the moving submarine’s location and path and identify the acoustic admissions of the submarine.

Project 2: Music Score Extraction from Audio Clips Using Gabor Transform and Gaussian Filtering

This project aims to reproduce the music score for the guitar in the two clips of the greatest rock and roll songs of all time - Sweet Child O’ Mine by Guns N’ Roses and reproduce the music score for bass and find guitar solo in 60 seconds clip of the song Comfortably Numb by Pink Floyd.

Project 3: PCA-Based Motion Analysis of Spring-Mass Systems Using Multi-Angled Video

This project aims to use the Principal Component Analysis (PCA) method to analyze the different datasets from videos of a spring-mass system recorded from cameras at three different angles. For each camera angle, we are provided with four separate cases. One of them is the ideal case, and the other ones have different variations, such as camera shakes. We will use PCA to compare and contrast the different instances.

Project 4: SVD Analysis and LDA-based Classification of MNIST Digit Images

This project aims to first perform an SVD analysis of the digit images from the MNIST data set, then use Linear Discriminant Analysis (LDA), support vector machines (SVM), and decision tree classifiers to identify individual digits in the training set and classify digits in the testing set.

Project 5: Dynamic Mode Decomposition of Video Streams for Foreground-Background Separation

This project aims to use the Dynamic Mode Decomposition method on the video clips to separate the video stream to both the moving objects in the foreground video and a static background.

Project 6: Evaluation of MALA and RWMH in Approximating Diverse Probability Distributions

This project explores the Metropolis Adjusted Langevin Algorithm (MALA) and compares it with the Random Walk Metropolis-Hastings (RWMH) in approximating Gaussian, multimode, and high-dimensional probability distribution settings. The project also studies the importance of optimal discretization step size in MALA.