kmzzhang / nbi

Package for Neural Posterior Estimation and Importance Sampling focused on Astronomical Applications
Other
34 stars 4 forks source link

Automatically select relevant training data from previous round #12

Open kmzzhang opened 1 year ago

kmzzhang commented 1 year ago

For each round > 1, calculate the probability of previous round training parameters under the current round surrogate posterior. Include all previous training data within e.g. 5\sigma probability into next round training data.