The Milky Way swarms with orbiting satellite dwarf galaxies of astounding diversity. Some galaxies continue to form stars while others stop and dim in brightness. In computer simulations, the evolutionary history of each dwarf galaxy that leads to these differences is known. Galaxies can lose gas and stop forming stars due to early exposure to stellar radiation (reionization), interaction with the hot gas of the host (ram-pressure stripping), or gravitational interactions with the host/dwarf galaxies (tidal effects).

We propose to use the characteristics of simulated dwarf galaxies from a suite of 20 Milky Way-like galaxies, each of which hosts ~100 dwarf galaxies, to train a supervised machine learning algorithm. The goal is to use only those dwarf galaxy characteristics observable in the present day as features to predict key aspects of their evolutionary history, such as the dominant mechanism by which they lose gas (classification), and/or the timescale of gas loss (regression). We will then apply the simulation-trained algorithm to the characteristics of our own Milky Way’s dwarf galaxy population, which have uncertain evolutionary histories. Identifying the likely histories of dwarf galaxies will help us to understand how galaxies form and evolve over time.

One selected candidate will receive a stipend via the DSI Scholars program. Amount is subject to available funding.

Faculty Advisor

  • Professor: Mary Putman (not recently updated)
  • Department/School: Astronomy
  • Location: Pupin Building
  • Galaxy formation through studies of gas flows and dwarf galaxies

Project Timeline

  • Earliest starting date:
  • End date:

Candidate requirements

  • Skill sets: Python (preferably), some statistics, some astronomy or willingness to learn
  • Student eligibility: freshman, sophomore, junior, senior, master’s
  • International students on F1 or J1 visa: eligible