Textbook Group meeting for Parr, Pezzulo, Friston 2022 .

right thank you for joining everyone it's 11:21 and we're in our first discussion for cohort 5 chapter 5 before we go to any specific questions where does anybody want to begin or just any anywhere chapter 5 related or adjacent that they want to just begin with any thoughts on the chapter overall or like where it sits in the book or just any anything else at the whole textbook level or chapter 5 level um yeah I just this is a little specific but so far I think this might be the chapter that goes the most into precision and that's kind of a I don't know that it's such a key concept it seems an active inference that I just find that really interesting here um I especially find the discussion of like like getting more into the neurobiological aspect of like different neurotransmitters corresponding to tuning like different precisions like dopamine policies and nor adrenaline Transitions and so on so yeah yeah where whereas often the material aspects of the brain are linked to the kind of estimate of the mean like with rate coding here it's a connection between neurotrans transmitters or or sometimes called neurom modulators because through Precision modulation they're influencing how different things happen of course recognizing that like even the same molecule because of the receptor being different and the downstream processing and the context being different like it's not even like one neurotransmitter does one thing even in one species brain so of course this is painting with a broad brush but still these are real research programs that have benefited from unifying different experimental results related to a neurotransmitter Under the Umbrella of like a model parameter specifically Precision parameters precision and and variance just kind of being like two ways to describe it like how thin or how wide it's like the same thing you're talking about and in chapter five being where any empirical evidence even comes in or any specific non-toy example comes in Oh chapter one overview low and high road chapter four just the math and the details of the generative model so here's the first time that we're even talking about a real specific system which gives a flavor of like the kind of experiments that people have done kind of framings that people take towards the mamalian nervous system human nervous system not to say that's the only system that active applies to or anything like that but just that that's where the most empirical work to date has happened so to provide just one more chapter in between chapter 4 describing the generative model and chapter six how to build your own chapter five is just like this kind of empirical Waypoint to show here's what some research programs involving chapters one through four look like okay now chapter six here's how you would do that in this system or in a different system this first paragraph has a important idea about the sparsity of the statistical model and then the sparsity of the anatomy of the system and in the brain where those two sparsities kind of coexist but having a sparser statistical model means that variables tend to only influence a few other variables and that can make things easier to compute or easier to interpret statistically now obviously you can simplify it or sparsify it to a point where you're missing out in P vectors then you've sparsified too much but to the right extent model sparsification makes it more computable more interpretable and the second sparsity is like the anatomical sparsity which is the fact that although there are a lot of connections in the brain with the dendrites and the neurons and all this and cells really are connected to a lot of other cells in the body they're not connected to all cells and so to the extent that not all things are connected to All Things most directly or in a given way that's the morphological sparsity and the kind of neuromorphic hypothesis is that the sparsity of the substrate has an interplay with the sparsity of the…