Using Machine Learning to Characterize Atom Chip Traps

NASA's Cold Atom Laboratory (CAL) is a compact experimental system that is deployed on the International Space Station. CAL experiments utilize atom chips to trap ultracold atoms, with a goal of generating and manipulating Bose-Einstein Condensates (BECs). Due to the compact design of the device, the magnetic field is more difficult to control and characterize than in a typical ultracold atom system. To help address these challenges, we have implemented a neural-network-based machine learning algorithm for analyzing ultracold atom experimental data. The goal of this method is to efficiently characterize the atom chip trap using time-of-flight data. We demonstrate that the neural network can be trained to learn the shape of the atom trap, using a simulated dataset. We will present results demonstrating the efficacy of this approach when applied to simulation data and the effect of different neural network architectures on the algorithm.