Machine Learning

This project was created using a Tensorflow LSTM which would recieve images from the client robot and make predicition for which turns it should make. The pictures on the client were taken using OpenCV2 and would be sent over as raw pixels in the form of a real-time web socket. The model learned to drive through a paper track on the ground and passed the statistical analysis proof.

The goal of this project was to see how effective a reinforcement learning type training model would be for financial management. Normal models rely on linear regression and predictions of the next price of a stock. Instead, this project sought to determine if policy gradient, a type of reinforcement model, may be effective for portfolio management.

The inspiration for this project came out of the recent rise in tele-medicine and doctors appointments hosted over the internet. As such, I wanted to see if there was a way to bring physical therapy to the online world to help people at home get the most out of their workouts. Current products already exist that allow users to track their joint movements as they do a series as exercises, though these solutions are often expensive. Instead, a machine learning technique would allow the users to just need their phone to complete their assigned workouts and gain assistance from the model.