The Large Hadron Collider accelerates protons close to the speed of light and then collides them with the goal of discovering new particles that will help us better understand our universe. The ATLAS Detector collects data on about one billion of these particle collisions, an amount of data that if stored on CD's would reach the moon and back twice a year. However, the vast majority of this data is low energy dijets, which have a low likelihood of leading to a new discovery. This project explores a novel method of filtering out the data using anomaly detection methods of machine learning. Specifically, various autoencoders, neural networks that compress and reconstruct data, were used to differentiate the levels of anomaly of each type of collision in the dataset.