This contains information about the experiment: Machine Learning X-ray cluster masses
Abstract: In this lab-experiment you utilize neural networks to infer the masses of galaxy clusters directly from observed/simulated X-ray images. You will work with convolutional neural networks who are a standard tool for successful inference on images. We emphasise how we can estimate the uncertainty associated with the neural network prediction using ensembles of neural networks and a likelihood based custom loss function. We discuss data augmentation methods to boost the neural network performance. We work with simulated X-ray simulations which were created to mimic eROSITA observations.
ML packages: Tensorflow, Keras
Further packages: Pandas
Input data: 3D arrays
Preparation: You are expected to work through this preparation document. Also do take a look at the documents on the lab day and the report which are linked below.
You are also expected to read through the following two guides.
Lab days: A guide for the lab days can be found here.
Lab report: A guide for the lab report can be found here.