## Random connection to maths

### Random connection to maths

I am surprised by the structual homology between a Markov Random Field and a cellular automata. It really looks like a stochastic/probablistic automaton.

http://homepages.inf.ed.ac.uk/rbf/CVonl ... 9/ORCHARD/

http://homepages.inf.ed.ac.uk/rbf/CVonl ... 9/ORCHARD/

### Re: Random connection to maths

Belief propagation looks very much like a cellular automata. (see this ising example)

### Re: Random connection to maths

Anyone fancy implementing a "backpropagation through time“ (BPTT) using CA as substrate?

- BlinkerSpawn
**Posts:**1942**Joined:**November 8th, 2014, 8:48 pm**Location:**Getting a snacker from R-Bee's

### Re: Random connection to maths

The temperature model (and possibly the belief propagation system too, although I don't know how that works) could probably be approximated on the VN neighborhood using states 1,2,...,n to represent temperatures of 1/n,2/n,etc., for some n.

Granted, you'd need a script to generate the necessary transitions and there'd be a lot of them, although you'd have access to the permute symmetry which would save some space.

Granted, you'd need a script to generate the necessary transitions and there'd be a lot of them, although you'd have access to the permute symmetry which would save some space.

### Re: Random connection to maths

My interest lies in exploiting message-passing to make fast inference algorithm. A more recent analogue seems to be dynamic bayesian network (DBN), it is remarkably similar to CA and an oscillator in CA corresponds to something like a stable cycle in the DBN. The difference between DBN and MRF is that DBN is embedded in time whereas MRF isn't (though I supposes possible). Anyway it'd be interesting to see how much of the result of CA is transferable to DBN, please comments if you guys have any thought/reference.BlinkerSpawn wrote:The temperature model (and possibly the belief propagation system too, although I don't know how that works) could probably be approximated on the VN neighborhood using states 1,2,...,n to represent temperatures of 1/n,2/n,etc., for some n.

Granted, you'd need a script to generate the necessary transitions and there'd be a lot of them, although you'd have access to the permute symmetry which would save some space.

### Re: Random connection to maths

So excited that I dropped my jaw due to this paper by Cosma Shalizi [1].

accompanying blog

1. Shalizi, C. R., Haslinger, R., Rouquier, J., Klinkner, K. L. & Moore, C. Automatic filters for the detection of coherent structure in spatiotemporal systems. 1–16 (2006). doi:10.1103/PhysRevE.73.036104

accompanying blog

**Reference**1. Shalizi, C. R., Haslinger, R., Rouquier, J., Klinkner, K. L. & Moore, C. Automatic filters for the detection of coherent structure in spatiotemporal systems. 1–16 (2006). doi:10.1103/PhysRevE.73.036104

### Re: Random connection to maths

I'm curious of how it works on the Turbulent phase cyclic cellular automata, where large-scale structure dominates.shouldsee wrote:So excited that I dropped my jaw due to this paper by Cosma Shalizi [1].

accompanying blog

Reference

1. Shalizi, C. R., Haslinger, R., Rouquier, J., Klinkner, K. L. & Moore, C. Automatic filters for the detection of coherent structure in spatiotemporal systems. 1–16 (2006). doi:10.1103/PhysRevE.73.036104

Still drifting.

### Re: Random connection to maths

This brain model looks very CA to me. https://www.youtube.com/watch?v=843G1WDnmAU