What are the best ways to improve a model #776
-
Hello, First, I would like to thank all of you for this awesome package, simulation-based inference is some sort of game changer in my research field and I would be glad to contribute to the diffusion of these methods. What I don't grasp yet are the best ways to improve the modelisation of the likelihood to derive quantitive results. I have various ideas in mind but don't know what I should start with:
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 3 replies
-
Hi there, thanks for starting this discussion! Overall, I think your questions are quite open-ended; I will post my opinions below, but these are not final answers and it would be great to hear the opinions of other other users here. Anyways... At which point getting more simulations in the dataset could help?In general, more simulations yields better results. In addition, more parameters will often mean that you will need more simulations, but it is very difficult to give answers such as "you need X simulations for a simulator with N parameters". However, very very roughly, I would say:
In order to identify if you need more simulations, I would recommend predictive checks and calibration checks Should I focus on getting better summary statistics or training an embedding net?General rule: if you can write down summary statistics (that are useful and you care about), I recommend using summary statistics (instead of using the raw data + embedding net). Using an In addition, the choice of summary statistics is key to performance of sbi, see also here. As spelled out in this tutorial:
Should I change the density estimator internal architecture (adding more flows or anything?)In many cases, |
Beta Was this translation helpful? Give feedback.
Hi there,
thanks for starting this discussion!
Overall, I think your questions are quite open-ended; I will post my opinions below, but these are not final answers and it would be great to hear the opinions of other other users here.
Anyways...
At which point getting more simulations in the dataset could help?
In general, more simulations yields better results. In addition, more parameters will often mean that you will need more simulations, but it is very difficult to give answers such as "you need X simulations for a simulator with N parameters". However, very very roughly, I would say: