Referring neuroscience, I remember when I was looking for some info on “biased random walk” (the behavior of E. coli) an article titled “Biased Random-Walk Learning: A Neurobiological Correlate to Trial-and-Error” came across me. It talked about the possibility that “random variation in synaptic strength” plays a role in the learning progress of neural networks.
The fulltext could be found by scholar.google.com with that title.
Best regards,
Bo
Gary Cziko g-cziko@UIUC.EDU Wrote£º
···
[from Gary Cziko 2007.05.02 16:33 GMT]
I sent the “E. coli’s control systems” conversation to my daughter, Anne-Marie Cziko, who is finishing her doctorate in neuroscience at the University of Arizona. She and her lab deals with micro aspect of nervous systems including genes, synaptogenesis and such.
She
responded:
These kinds of things have been known about in biological systems for years! All signaling networks have robustness built in that can handle a huge deal of disturbance. We talked about this for several weeks in one of my graduate student classes.
She will send me some references on “synaptic homeostasis” as an example (those impatient can Google “synaptic homeostasis” and get 881 hits.
–Gary
P.S. It would be interesting for someone to write a paper or a book about all these biological control systems so nonspecialists could understand them, and then show how similar systems might even be involved in animal and human behavior!
ÇÀ×¢ÑÅ»¢Ãâ·ÑÓÊÏä3.5GÈÝÁ¿£¬20M¸½¼þ£¡
Bill Powers powers_w@FRONTIER.NET Wrote:
Most interesting, Bo. The mathematics is far beyond me but the principles seem closely related to my E. coli method of reorganization. I notice that several papers seem to be using this method to reorganize perceptual input functions. Also, one paper applies it to the grazing of animals, which is appealing since it is like the E. coli way of finding food! Perhaps Richard Kennaway, who has nothing else to do with his time (sorry, Richard) will look at this material.
In the field study so-called “sensorimotor integration” or “pattern generation”, more and more attentions are paid to some random search processes that provide a system with adaptive characteristics. These methods, like evolutionary algorithm and biased random walk, are said to be good at tracking those easy changing best
fit point. Math helps us in a different perspective.
I recommend a re-reading of Chapter 14 because near its end there appears the ghost of that chapter on “An experiment with awareness” that the editors decided to leave out of B:CP. It’s all about how awareness directs reorganization, so that a probable principle of therapy would be “change demands consciousness from the point of view that needs changing.” I think that’s the earliest allusion to the method of levels in print.
I appreciate your ideas that desire and error cause the change. The common explanation with reinforcement just describe the phenomenon again without looking into it.
That was a good find, Bo. Also, I did’t know about Scholar.Google.com – very useful.
Thanks. See you in July?
See you then: )
Best regards,
Bo
···
ÇÀ×¢ÑÅ»¢Ãâ·ÑÓÊÏä3.5GÈÝÁ¿£¬20M¸½¼þ£¡
I will look into it, thank you~
Best regards,
Bo
Dag Forssell csgarchive@PCTRESOURCES.COM Wrote:
···
[From Dag Forssell (2007.05.05.1330 PST)]
At 08:47 PM 5/4/2007, you wrote:
Please give this newbie the references for LCSII and MOL.
Ted
Check out www.livingcontrolsystems.com. Check it again in a month or three. I am updating and upgrading.
Best, Dag
ÇÀ×¢ÑÅ»¢Ãâ·ÑÓÊÏä3.5GÈÝÁ¿£¬20M¸½¼þ£¡