Finite state systems

Share on FacebookShare on Google+Email this to someoneTweet about this on Twitter

HOW E. COLI BACTERIUM GENERATES SIMPLICITY FROM COMPLEXITY

“In a surprise about E. coli that may offer clues about how human cells operate, the PNAS paper reports that only a handful of dominant metabolic states are found in E. coli when it is “grown” in 15,580 different environments in computer simulations.”

“When it comes to genomes, a great deal of complexity boils down to just a few simple themes,” said Bernhard Palsson, a professor of bioengineering at UCSD’s Jacobs School of Engineering and co-author of the study, which was made available online Dec. 15. “Researchers have confirmed the complexity of individual parts of biochemical networks in E. coli and other model organisms, but our large-scale reconstruction of regulatory and metabolic networks involving hundreds of these parts has shown that all this genetic complexity yields surprisingly few physiological functions. This is possibly a general principal in many, if not all, species.”

Similar principals have broad application.

1) Thousands of elements interacting in nonlinear dynamic feedback systems.
2) Evolutionary competition with survival of the fittest.
3) Subsets of elements operating together to maintain a specific state. Optimized to support that state.
4) States associated with successful survival strategies.

Applications in low-level neural circuits, high-level thought patterns, knowledge domains, belief systems, and social organizations.

5 Comments

  1. Sounds like a hidden markov model to me.  
     
    We use them in the analysis of wireless systems which again have thousands of interacting nonlinear systems yet have emergent states.

  2. Jody, I?d like to hear more. 
     
    How many and what types of emergent states do you see? How large is the wireless network and what is the predominant network topology? What types of traffic, telephony or data? Are you estimating point-to-point offered traffic from connection and blockage statistics on the wireless network links?

  3. Let me give an example – a colleague of mine is doing his dissertation on applying hidden markov models (HMM) to to cellular systems (so topology is predominately star, though soft handoffs blur the topology for his application) to predict call drops. 
     
    Answering the other questions before continuing to a broader description, in his work, the traffic in the network is a mix of voice, video, and data (read as text messages and file downloads) and traffic patterns are estimated via measurements and Motorola’s massive multi-cell simulator (he’s on a Motorola UPR). It is my understanding that the simulator incorporates thousands of simulated devices some of which have little effect on a particular call, some of which have a greater effect.  
     
    Using a HMM, he’s able to show that a call passes through sevenish states with the exact number of states and exact transition probabilities a function of the multipath profile of the device in the call and the load and positioning in the network (not a lot of variation though). (I would be able to be more specific, but I missed his prelim defense and am relying on second hand reports.) 
     
    The upshot being, he can estimate the probability that a call will shortly transition into the “dropped-call” state. Eventually, the network would reallocate resources to reduce the likelihood of a dropped call. Of course that changes all the transition probabilities, and I’m not certain if the effect of the implicit network adaptation recursion has been worked out yet.

  4. HMMs have many applications in genomics.

  5. Jody, ?he’s able to show that a call passes through sevenish states with the exact number of states and exact transition probabilities a function of the multipath profile of the device in the call and the load and positioning in the network (not a lot of variation though).? 
     
    Thanks for that info. With a dissertation that interesting, I?m sure your friend has a promising career. 
     
    ?I’m not certain if the effect of the implicit network adaptation recursion has been worked out yet.? 
     
    I can imagine the difficulty; each potential adaptation requires another network simulation.

a