Although probably late in this field, I have recently come across the topic, sourced/starting from an old paper by one of the main actors in this area, Gheorghe Paun ("An impossibility theorem for indicators aggregation"), who I then followed through to learning about Membrane Computing. I am planning to buy the introductory book on the topic, also authored by Paun, but not right away. I was wondering if there are individuals who could speak to the how and why, and maybe work on applications in this space.
I have worked with it [1], in particular with a special class of it, MP (Metabolic P).<p>In a nutshell, it is a theoretical model <i>inspired</i> by biology, but which produces many interesting results, as super-Turing computational power.<p>I am not sure whether it is still, but the reference on web for the subject was <a href="http://ppage.psystems.eu/" rel="nofollow">http://ppage.psystems.eu/</a> .<p>Besides Păun, other big names in the topic are Perez-Jimenez, Gheorghe and Manca.<p>As far as I am concerned, there are little practical application of it, even though the theoretical potential is considerable.<p>I tried to focus on more day-to-day application of MP [2], and I managed to get some nice results, but further advance was limited by external forces.<p>Do I recommend to buy an introductory book on the subject? No, mostly because I think the material online from the people I mentioned above, plus their publication, is more than enough to acquire the knowledge.
Besides, I bet the book is published by Elsevier, which usually mean it is a simple collection of papers, except the fact I am not fan of that publishing house.<p>If you have any more direct question, feel free to contact me—hopefully I can be of some help.<p>But, I have the impression that membrane computing is still a very academic topic disconnected from the engineering world.<p>If it is what you like, then it might satisfy you. :)<p>[1]: <a href="https://ricardo.guiraldelli.com/research.html#doctorate" rel="nofollow">https://ricardo.guiraldelli.com/research.html#doctorate</a><p>[2]: <a href="https://ricardo.guiraldelli.com/research.html#available-material" rel="nofollow">https://ricardo.guiraldelli.com/research.html#available-mate...</a>
<a href="https://en.wikipedia.org/wiki/Membrane_computing" rel="nofollow">https://en.wikipedia.org/wiki/Membrane_computing</a><p><a href="https://en.wikipedia.org/wiki/P_system" rel="nofollow">https://en.wikipedia.org/wiki/P_system</a><p><a href="https://en.wikipedia.org/wiki/Cell_membrane" rel="nofollow">https://en.wikipedia.org/wiki/Cell_membrane</a><p>Huh. Wasn't familiar with this before now. Looks very interesting though.
In a somewhat-related area, Alan Kay viewed object-oriented programming as cells (objects) and their transport proteins (messages). Like biology, the system is a big ugly mess, but it's made up of tiny self-contained units that can handle themselves. The actor model continues this idea, but usually drops the connection to biology.
The claims of "solve NP problems in linear time" appear to come with a giant gotcha: they involve allowing the number of membranes to grow exponentially with the problem size! For instance, see:<p><a href="https://www.cs.auckland.ac.nz/research/groups/CDMTCS/researchreports/102paun.pdf" rel="nofollow">https://www.cs.auckland.ac.nz/research/groups/CDMTCS/researc...</a><p>This seems like a huge problem for any electronic representation, since most any computational approach will deliver "super-Turing" results if you allow the machine to grow faster than the problem...
This seems to be a special case of a computational fabric.[1] If you reduce computation to the barest minimum, you get down to a cartesian grid of 4 input 4 output Look Up Tables (LUTs). If you clock them in alternating phases, like colors on a checkerboard, you avoid all race conditions, and get deterministic general purpose computing which can run this type of algorithm.<p>I was reading George Gilder back in the 1980s when that idea hit me, and its been in the back of my brain ever since.<p>[1] <a href="https://en.wikipedia.org/wiki/Fabric_computing" rel="nofollow">https://en.wikipedia.org/wiki/Fabric_computing</a>
Similar to VM emulation, the problems with electronic or Turing simulation of chemical and biological processes are many. Simulating a sea of neurons or organelles doesn't tend to scale well without using actual organic processes.<p>I wouldn't bother without a specific need that can be accelerated with biochemistry directly. Turing completeness can't be improved on.
Luca Cardelli, who is quite famous in formal methods, did a lot of work in the area during the early 2000s: <a href="http://lucacardelli.name" rel="nofollow">http://lucacardelli.name</a>