Constraining the Complex Relationship Between Galaxies and their Dark Matter Haloes with Machine Learning

该项目将利用一类被称为稀疏回归方法（SRM）的机器学习算法来提取星系的各种属性之间的强大关系，它们的气体环境和它们的宿主暗物质晕，这些晕是在星系群的宇宙学流体动力学模拟中形成的。大量的星系参数关系可能有潜力被预测的模拟（如形成时间，自旋，合并的历史），但很难通过直接的手段来量化。这就是SRM的好处，因为它是一种（与其他机器学习技术不同）丢弃不需要的自由参数并仅从系统的状态描述中有效地提取物理系统的“控制”方程的方法，而不需要对相关物理的详细事先理解。这也将允许星系群被“画”到仅暗物质的模拟上，与通过称为晕建模的过程的完整重子模拟相比，这相对便宜。

The project will utilize a class of machine learning algorithm known as sparse regression methods (SRM) to extract robust relationships between the diverse properties of galaxies, their gaseous environments and their host dark matter haloes, formed in cosmological hydrodynamical simulations of the galaxy population. A vast quantity of galaxy parameter relationships may have the potential to be predicted by the simulations (such as formation time, spin, merger history), but are difficult to quantify through direct means. This is where the benefit of SRM is presented clearly, as it is a method that (unlike other machine learning techniques) discards unneeded free parameters and efficiently extracts the "governing" equations of physical systems from state descriptions of the system alone, without a need for detailed prior understanding of the relevant physics. This would also allow for galaxy populations to be "painted" onto dark matter-only simulations, which are relatively inexpensive to generate compared to full baryonic simulations through a process known as halo modelling.