Detecting Exoplanets Through Transits

For my Supercomputing project, I have decided to analyze various collections of data (put forth by NASA Citizen Scientist Programs) for the cause of finding an undiscovered exoplanet—that is, a planet outside our own solar system. I propose to use Machine Learning to search this data for any signs of a star suddenly dimming in brightness for a small period of time and then returning to its normal amplitude of light, an occurrence called a light curve. If I detect such a thing, I will be able to assume that the occurrence was due to a transit, in which an exoplanet that neither the human eye nor advanced telescopes can detect crosses its home star during its orbit. The shape of the planet crossing over its star will lessen the amount of light waves that reach Earth, thus making the star “dimmer.” This technique has been used by professional astrophysicists many times to help them discover many of today’s known exoplanets. Thus, I believe transit data is a reliable foundation for my project. To train my machine, I will use the Pytorch neural network, which I will train with model data that I will generate. Then I will apply it to data from NASA’s Exoplanet Watch and perhaps from my own telescope.