Predicting the stability of planetary systems

Planets form in protoplanetary disks of gas and dust, which damp their motions onto planar, circular paths. The final stage of planet formation occurs after that gas dissipates, setting off a phase of violent impacts, which determine the final masses and orbital configurations we see today. Our group has been working on understanding the chaotic dynamics leading to these instabilities and to quantitatively predict what subset of orbital configurations could survive to the present day. This a key question if we want to connect theoretical models of formation in the protoplanetary disk to the observational data we have at billion-year ages.

A few years back, our group combined our partial understanding of the underlying dynamics with machine learning techniques to train a model capable of predicting stability of planetary configurations over a billion orbits (see attached publication). Since then, we have made substantial theoretical progress on understanding the dynamics. Our goal for this summer is to exploit this new dynamical understanding to train an improved machine learning model. 

For your essay, please write a short paragraph on a programming project you're proud of, what the biggest challenges were, and how you went about debugging/solving them? Doesn't have to be a major project, can be something from CS 5.