I encourage those without a background in physics to first check out the Particle Physics Basics section to better understand my research. I tried to write as jargon-free as possible, but still, much jargon is unavoidable.

Presently, I am involved in a search for Beyond Standard Model Physics in proton-proton collisions at the Large Hadron Collider (LHC). The LHC, located at CERN in Geneva, Switzerland, is the world's largest particle accelerator. The LHC collides alternating beams of protons at nearly the speed of light, resulting in highly energetic collisions that allow scientists to explore exotic particle states not seen in normal conditions. The Compact Muon Solenoid (CMS) detector is one of two multipurpose particle detectors designed to analyze almost every aspect of these collisions. A worldwide collaboration of over 5000 scientists work together with CMS to expand our knowledge of fundamental particle physics.

Within the CMS collaboration, I am working on Machine Learning techniques to search for new particles not described by the Standard Model- the highly-successful theoretical framework for the most elementary particles in nature. Specifically, I am searching for Higgs Boson-like particles that decay to photons. 

To perform this search, we begin by making images from the energy deposited by photons in the detector. Because the portion of the detector responsible for detecting photons- the Electromagnetic Calorimeter- is composed of many individual crystals, we can take the energy deposited into each crystal as a pixel value, and produce images for each event.

The images will appear different for events in which a single photon, two photons, or more complicated "hadronic" objects strike the detector. The images show clear differences between each type of event. I am using Machine Learning in the form of a Convolutional Neural Net (CNN) to "view" thousands of these images and decide which type of event it was simply from the image. The CNN works by "training" on pre-labeled images, in which the event type is known. The algorithm then "learns" the key distinctions between each type of image and makes predictions for unknown images. Currently, the CNN is able to distinguish these images with very high accuracy!

While we are still in the development phase, it is promising that the CNN will be a powerful tool for this analysis and others within the CMS collaboration and beyond. Publication information will be posted when available.