These slides show the independent project I worked on in creating an algorithm that takes a mask of an image and fills in that part of the image with other parts of the image. A very fun, but challenging project.
A paper about my implementation of the reinforcement TD learning algorithm. This is a very dynamic algorithm because it reuses its inputs to make deeper inferences of future inputs to predict output. Compared to a supervised algorithm, this algorithm has more flexibility using its lambda parameter which sets a level of expectation from its previous iterations. Truly an amazing algorithm.
I took footage from the a 2012 debate presidential debate and ran a particle filter to track the face and hand of Mitt Romney. It was a lot of fun, but difficult setting the parameters once noise was applied.
Wrote analysis on two datasets using several kinds of unsupervised algorithms. One dataset came from pictures taken in a national park of foliage while, the other was a facial recognition dataset. Both datasets had clustering algorithms such has K-means clustering, Expected Maximazation. Feature selection algorithms as well were used in the analysis.
This is a github repository for my full stack android application. Server sides runs through the Spring environment. By using the Spring framework, I was able to fluidly implement REST implementations to allow client/server communication. Applications included uploading media to a server through a secure client using Oauth 2.0 technology and using a SQL database through a content provider framework. This project was done by myself. The idea was to create an app that you can share images and like other peoples images.