trashBinPoster.pdf
Gradient Descent
In this project, I simulated the process of optimizing a vehicle's suspension system using gradient descent algorithms with and without momentum. Given a function of two parameters k and c (representing spring stiffness and damping coefficient), the algorithm finds the optimal values of each parameter that minimizes the function, representing the suspension system with the least disturbance.
Fashion-MNIST
(Neural Network)
In this project, I designed a neural network capable of classifying images in the Fashion-MNIST dataset with a final accuracy of 85.7%. This project taught me all of the necessary aspects of creating, training and testing neural networks.
Nonlinear Regression
In this project, I compared the accuracy of Fourier and exponential basis functions in approximating the angle (theta 1) of a double pendulum system over time. I then compared the results of each basis function to understand why the accuracy varied between methods.
Autoencoders
In this project, I simulated a common engineering problem of analyzing measured data to determine which part of a system is malfunctioning. Using data from two different suspension systems, I created, trained and used an auto encoder to embed the data into two separate clusters which I was then able to extract and plot. This allowed the faulty suspension system to easily be determined.