Pearl blankets vs. Friston blankets
- Papus79
- Posts: 1800
- Joined: February 19th, 2017, 6:59 pm
Pearl blankets vs. Friston blankets
http://philsci-archive.pitt.edu/18467/1 ... ankets.pdf
I'd be curious if anyone else has looked into this and gathered their own thoughts on this? I personally find the Markov chains fascinating as they offer some type of formal inroad into processes, and it doesn't surprise me that there'd be something of a gold rush on it in the research communities - both in biology and computer sciences.
I'm still somewhat early in reading this paper and my mind could be changed on this but my sense is that when you're in a hot area and you're getting a lot out of it - it might take a while before you start feeling comfortable laying down 'no go' zones or at least stipulating new ones. That could be part of why they're leaving map and territory distinctions a bit blurry. The paper above is only 47 pages long so I think its something I could cover end to end in the next few days.
- Terrapin Station
- Posts: 6227
- Joined: August 23rd, 2016, 3:00 pm
- Favorite Philosopher: Bertrand Russell and WVO Quine
- Location: NYC Man
Re: Pearl blankets vs. Friston blankets
- Terrapin Station
- Posts: 6227
- Joined: August 23rd, 2016, 3:00 pm
- Favorite Philosopher: Bertrand Russell and WVO Quine
- Location: NYC Man
Re: Pearl blankets vs. Friston blankets
- Papus79
- Posts: 1800
- Joined: February 19th, 2017, 6:59 pm
Re: Pearl blankets vs. Friston blankets
I don't know what your familiarity is with Markov chains but I've been interested for a while before this in how those map dynamic equilibrium in complex systems. It appears that a Markov blanket might be a way of framing the cradle of inputs that those rest in and so they'd lend themselves well to troubleshooting problems within complex systems. That's part of why I can see a lot of the intuitions that Karl Friston et al are voicing, particularly with what kinds of inroads this kind of math that provide in understanding what goes right or wrong in biological ecosystems, political systems, even JIT supply chains or traffic infrastructure, and how to improve our maintenance, building, and perhaps when or if we have to nudge an ecosystem or social system in one direction or another we can do it with fewer ugly second and third order consequences.Terrapin Station wrote: ↑April 30th, 2021, 7:08 pm In general, by the way, I'm even rather skeptical of probability, but I'm especially skeptical of Bayesian approaches to probability.
- Terrapin Station
- Posts: 6227
- Joined: August 23rd, 2016, 3:00 pm
- Favorite Philosopher: Bertrand Russell and WVO Quine
- Location: NYC Man
Re: Pearl blankets vs. Friston blankets
I'm somewhat familiar with Markov chains, but what I've always been curious about since becoming familiar with Markov chains but I've yet to explore very much is the extent to which real-world empirical verifications in a variety of contexts bear out the probability predictions of Markov chains.Papus79 wrote: ↑April 30th, 2021, 7:18 pmI don't know what your familiarity is with Markov chains but I've been interested for a while before this in how those map dynamic equilibrium in complex systems. It appears that a Markov blanket might be a way of framing the cradle of inputs that those rest in and so they'd lend themselves well to troubleshooting problems within complex systems. That's part of why I can see a lot of the intuitions that Karl Friston et al are voicing, particularly with what kinds of inroads this kind of math that provide in understanding what goes right or wrong in biological ecosystems, political systems, even JIT supply chains or traffic infrastructure, and how to improve our maintenance, building, and perhaps when or if we have to nudge an ecosystem or social system in one direction or another we can do it with fewer ugly second and third order consequences.Terrapin Station wrote: ↑April 30th, 2021, 7:08 pm In general, by the way, I'm even rather skeptical of probability, but I'm especially skeptical of Bayesian approaches to probability.
-
- Posts: 10339
- Joined: June 15th, 2011, 5:53 pm
Re: Pearl blankets vs. Friston blankets
- Papus79
- Posts: 1800
- Joined: February 19th, 2017, 6:59 pm
Re: Pearl blankets vs. Friston blankets
I actually saw this used in quite different ways.
It seems like one analogy, the difference between Markov boundaries and Markov blankets, is that a boundary is the minimum set of states or variables where to drop below that threshold loses information. Where I get a bit confused is when they say 'useful' information for Markov blankets, possibly a bit like added capacity in the context of disequilibrium systems?
Friston actually seemed to use this in a particular manner with living systems, talking about what I think a lot of people term the inside/outside or subject/object boundary where you have information going out, information coming in, that that barrier facilitates certain kinds of gating and exchange that - going with the 'Free Energy Principal' act as aids not only in terms of protecting the interior states but also aiding in reduction of uncertainty or wasted / free energy.
I think the thing that was really interesting to me, at an entry level, was seeing that if you had an equal exchange of different effects between modes then enough circulations of those exchanges lent itself to a steady state (ie. crunching a given matrix enough times). That seems to describe what happens when you have plants, animals, etc. with relatively fixed characteristics, fixed amounts of prey, you get multiyear predator/prey cycles, if something can gestate long enough you get things as odd as cicadas whose cycles are set to prime numbers, I'm not sure how far down the drive for 'survival' can be mapped but at least in biological life it seems like it almost 'grids' relationships. I also think back, on a much different level, to things like Slate Star Codex's 'Meditations on Moloch' and how our culture really gets electrified by basins of attraction where those basins may not be rational but they're statistical and it's a bit like if you aren't in those 'right places' or tuned to chasing them, being early to them, etc., you can fall farther and farther behind.Steve3007 wrote: ↑May 25th, 2021, 5:15 amI am interested, though, in the general theme of using probability and statistics to arrive at effectively deterministic end results. Possibly this is where Markov chains have applications to thermodynamics and statistical mechanics? It's always been interesting to me that laws of Nature with rock solid reliability, and which seem to say useful things on subjects like non-reversible physical processes, are statistical. And therefore they are only applicable to systems with large numbers of parts.
Friston was talking about steady states and mentioned circular motion/orbit as one kind of steady state and plasma as another, and where life seems to come into play is where you have these orchestrated states of... I guess you could say collapsed or 'bundled' disequilibrium where you get organized flow.
For determinism, I don't if I've ever felt comfortable trying to pick that off from physical relationships but I've much more often felt comfortable looking at relationships in time (gotten into the weeds with a lot of people in free will threads) where for me if time is as linear as it presents itself then you should find no more high fidelity replica (perfect or near-perfect) as replaying some short slice of time in your life (a few seconds to a few minutes) because the starting inputs and states should - in theory - be identical every time, the environment unfolding around you should be identical every time, so there doesn't seem to be anyplace to than on 'some' kind of optimal path (what may or may not look optimal to an outside observer but the outside observer can't see all of your internal states).
- Papus79
- Posts: 1800
- Joined: February 19th, 2017, 6:59 pm
Re: Pearl blankets vs. Friston blankets
https://www.youtube.com/watch?v=d4KrWxrP-T0
- Sculptor1
- Posts: 7148
- Joined: May 16th, 2019, 5:35 am
Re: Pearl blankets vs. Friston blankets
A human conceit wrapped in a fantasy wrapped in an abstraction.
No such things exist out side the human imagination, so you cannot expect any of it to make any more sense than any other fantasy from Acromantula to Zoony the Space dog.
- Papus79
- Posts: 1800
- Joined: February 19th, 2017, 6:59 pm
Re: Pearl blankets vs. Friston blankets
2024 Philosophy Books of the Month
2023 Philosophy Books of the Month
Mark Victor Hansen, Relentless: Wisdom Behind the Incomparable Chicken Soup for the Soul
by Mitzi Perdue
February 2023
Rediscovering the Wisdom of Human Nature: How Civilization Destroys Happiness
by Chet Shupe
March 2023