32nd Annual Student Research Conference
Parallel Computation: Efficient Physics Simulations with Tensorflow
Katey J. Thompson
Prof. Timothy Wiser, Faculty Mentor
TensorFlow, Google’s machine-learning APO, has the potential to be a useful tool for physics simulation for systems with large numbers of initial conditions. Most relevant for our purposes, however, is that TensorFlow does parallel computation using GPUs. We have been working on simulations of basic physical systems in order to investigate the capabilities of the API. In the future, we would like to simulate a large number of dark matter particles in the solar system. Since the mass of dark matter in the solar system is small compared to the total mass, the internal interactions are effectively negligible. Ideally, TensorFlow will allow gravitational calculations to be performed simultaneously (i.e. more efficiently than other APIs are capable of).
Keywords: TensorFlow, physics, simulation, programming
Topic(s):Physics
Presentation Type: Oral Paper
Session: 210-5
Location: MC 211
Time: 11:15