Habit learning and precision

[PetrAnokhin]

Hello!
I couldn't find implementation of habit learning (updating of prior over policies E) and precision(gamma) updating through betta prior.
Am I missing it? And if not, is it planned for future releases?
Thanks!

[conorheins]

Hi @PetrAnokhin ,

Thanks for your message -- that is a good observation, you are correct we have not implemented habit learning (updating of prior over policies pE) nor precision updating of gamma through the beta prior. We have planned it for future releases! I will bump it up in the priorities list, thanks to your comment.

Apologies that this feature hasn't been available sooner :/

[PetrAnokhin]

@conorheins Thank you!

Looking forward to it!

[conorheins]

I forgot to mention that @daphnedemekas and I implemented a version of policy prior updating in a version of pymdp that we used in a separate project, see here

So in case you need to implement this feature sooner than we can add it to the official pymdp repo, you could look at that for inspiration

[XuLei9292]

Dear @conorheins,
I just start working on the topic of active inference, I wonder if it is available tuturial codes for constructing a hierachical model to update gamma?

[conorheins]

Hi @XuLei9292 ,

Unfortunately we have not implemented updating gamma in pymdp. However, there is an example of a hierarchical model you can find here: https://colab.research.google.com/drive/1PHVe4cr781YJyVDq4imozx-wbrleU1SC?usp=sharing

On a related note @xugami: I received an email notification that you posted a message to this thread, but actually I haven't seen it here -- so maybe you deleted it? But in the email notification you outlined (in pseudo-code) a promising algorithm for the steps needed to gamma in pymdp. If it's alright, could you share that here again, in case it's helpful to those like @PetrAnokhin or @XuLei9292 looking for ways to update gamma?