LFITaskForce / BerkeleyMeeting

Bay Area Likelihood-Free Inference meet-up at the Berkeley Institute for Data Science
https://lfitaskforce.github.io/BerkeleyMeeting/
1 stars 3 forks source link

[TUTORIAL] Amortized LFI of cluster masses with TensorFlow Probability #6

Open EiffL opened 4 years ago

EiffL commented 4 years ago

Amortized LFI of galaxy cluster masses with TensorFlow Probability

Learn to use TensorFlow Probability to get posterior mass estimates of cluster masses using dynamical mass measurements.

Contacts: @EiffL Participants:

Detailed description

This tutorial illustrates how to build an amortized inference model for estimating galaxy cluster masses using TensorFlow Probability (TFP). It covers the definition of a conditional density estimator using Keras, understanding the impact of the proposal distribution used to build the training dataset, and learning how to get proper Bayesian posteriors.

Ressources: colab notebook