search

ReactiveBayes / RxInfer.jl

Julia package for automated Bayesian inference on a factor graph with reactive message passing
https://rxinfer.ml/
MIT License
278 stars 23 forks source link

Examples need adjustments for new GraphPPL #242

Closed bvdmitri closed 7 months ago

bvdmitri commented 7 months ago

Migration guide

Basic examples

Advanced examples

Problem specific

Nimrais commented 7 months ago

@bvdmitri, I would like to fix Infinite Data Stream and Gamma Mixture.

HoangMHNguyen commented 7 months ago

@bvdmitri I choose GPSDE and RTS vs BIFM

Sepideh-Adamiat commented 7 months ago

@bvdmitri I choose the Mountain Car and Hidden Markov Model.

wouterwln commented 7 months ago

On the Mountain Car example @Sepideh-Adamiat , it might be worthy of a revision in general. The current example is, I think, an example of @ThijsvdLaar 's 2019 paper on "Simulating Active Inference Processes by Message Passing", and you can see the ideas of the paper explicitly named and coded as such. However, as an example of RxInfer, the main goal is not to demonstrate the ideas of a paper, but to demonstrate how to make an agent operate in a simple environment. Therefore I think we might need to adopt "standard" coding practices in the Mountain Car demo too (instead of creating a function that returns a callback function that reads/writes to global variables), to make clear what code specifically belongs to RxInfer and what code is just there to whip up the environment. Let me know what you guys think

Raphael-Tresor commented 7 months ago

@bvdmitri I would like to fix the 'Probit model' example and "Chance-constraints"

mhidalgoaraya commented 7 months ago

@bvdmitri I choose the Kalman Filtering and Smoothing example

bartvanerp commented 7 months ago

Awesome work @wouterwln and @bvdmitri ! Migration was peanuts!

albertpod commented 7 months ago

@wouterwln as for Mountain Car. It's a great idea. We have more people interested in this particular example, but I agree that the way it's coded up isn't pretty. Mindful of that it needs to change, I am not sure what would be the best way. AFAIR, you had an implementation of this example with RxEnvironments. Could it be something to consider?

Sepideh-Adamiat commented 7 months ago

@wouterwln Regarding the mountain car example: I also agree that the current version is confusing and the coding should be improved. I can start working on that. I will try to make a PR in the upcoming weeks.

bvdmitri commented 7 months ago

Thanks everyone! It is super helpful! There is one example left, but I'm going to close the issue, as it becomes a bit difficult to manage different branches with different fixes together with the examples branch.