Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

Hello, it is January 25th, 2023 and we are in Cohort 2 of the Par et al textbook. We're in meeting 13. Today we are going to jump back in, start our second interval of activities. There are a few ways we can go. We can just review structural aspects of the the platform and accommodate anyone's suggestions or notes on some things that could be interesting to develop with in this one or in a side little sandbox. Then we will look to the textbook itself and we will look at where the first five chapters got us and where we're going with the second half of the book which has a different form and function than the first half of the book. So first off, anyone want to just add any general comment how now that it's 2023 where are they in their act-inf learning and application journey? How is our niche similar or different than it was in September through November when we worked through the first five chapters? Any thoughts or we can continue on? Just wanted to leave that space. Okay, feel free to raise your hand or just unmute or put in the chat or anything like that. Broadly, the layout is the similar. Textbook group overview has information on our live meetings and the layout. Textbook section itself contains different types of things that are in the textbook such as the equations which have all been brought out with images and page at least but there's a lot of improvements that we could make to the descriptions and tagging. Similarly with figures we have all figures brought out but there's a lot of contributions that can be made to descriptions and tagging. Boxes and tables similar. Code implementations. I'll add a link to Blockference so if anybody has any of them can be able to do that, which is one thing that we've worked on the last few months, we have about 40 or 50 implementations of act-inf in Python, C++, Julia, MATLAB. So accessible from here and we can for those who want to start getting more into the code and I'll just show one more piece that we worked on earlier today in the Blockference meeting. In ModelStream 7.2, a notebook was used. An epistemic contextual multi-armed bandit simulation was provided by Connor Heinz. Just works perfectly, no tweaking is needed. Uses act-inf loop, defines, etc. And what we did today was we took the act-inf ontology and we then said how is this set up in the contextual bandit? How do they set up B? Okay, here's how B is set up. One can click this and then in a plain text paste in the whole script. So this is not the only way that we're going to be working towards modeling but as the second half of the textbook is about building a generative model and the process, the recipe, we have like multiple functions in this cohort two as we go through chapter six or ten. One function will be to, to the extent that we want to, develop our own generative model and simulation just didactically, just to have something that we increment along to help us understand, or if there's some outcome that people want to pursue. Research or application oriented, like second half of the textbook is the time to do it and even relative to last September, we have significantly better tools. So there's totally an approach to go from a natural language understanding of these ontology terms to how to do it in PiMDP. And it's pretty reasonable that somebody who wants to pursue this direction, we could just add another column, copy it over, make the tweak. And there's a few other pieces, but a lot of that is going to come up in chapter six. In chapter notes and questions, we have the chapters themselves and a detailed table of contents. And then each chapter has subset view of the questions. So if you have a question within a given chapter or you want to help improve and curate the discourse, please just like add a row in the questions section or just make improvements to the answers and discourse. So we'll continue to work with questions as well as whatever notes people add to just what or anything else they want to…