[Martin Taylor 2013.12.19.00.44]
[From Rick Marken (2013.12.18.1530)]
Martin Taylor (2013.12.17.23.28)--
MT: There are three separate issues here:
(1) What did Bill Powers define as "e-coli reorganization". For that I
refer you to Chapter 7 in LCSIII, particularly pp66 (in the review version
-- I think my paper version must be at my cottage) and Demo 7-1. He
emphasises the fact that e-coli keeps going in the same direction when
things are getting better, but tumbles quickly when they are getting worse.
He discusses how just changing the rate at which changes are made got him
nowhere until he thought of the directional constancy of e-coli that is
going "the right way". So rate changes are indeed only half the story that
is e-coli reorganization.
RM: I can't find that in my copy of LCS III. But I think (having
writtin some e.coli simulatons myself -- see
http://www.mindreadings.com/ControlDemo/Select.html) that directional
constancy (or lack thereof) is a side effect of the rate of tumbling
I think you have that backward, at least as Bill thought about it. It was the directional idea that solved his problem. You can't even have the concept of a tumble if you don't first have a direction. Here is the relevant section of LCS3, pp 65-66 of the review version PDF (retyped by me, so there may be typos that are not in the book). The elisions are irrelevant side comments.
--------quote Powers LCSIII pp65-66-------
P65: If there is some condition in the organism that causes changes in
behavioural organization to occur, and if removing this condition stops
those changes, then of course the organization of behaviour will remain
as it was just before the condition was removed. It will look as if
removing that condition is reinforcing because that removal, or the
positive occurrence of whatever removed it, seemed to strenghten the
tendency to produce that behaviour from then on. What is really
happening, according to reorganization theory, is that the tendency to
change the behaviour is reduced. The idea of reinforcement is the
opposite of the process that I called reorganization.
While this was not quite a workable model, it did contain the seeds of
one. It showed that if an error signal in some sort of very basic
control system could cause outputs that randomly altered the
organization of behaviour, this reorganization would continue until a
behavioural organization was produced that would bring the error signal
to zero. ... I more or less abandoned this model because on second and
third thought it just didn't seem efficient enough.
In 1980 Daniel Koshland published a little book called "Bacterial
Chemotaxis" (1980) which provided the missing feature of the model of
reorganization. ... When all these flagellae spin at the same time in
the same direction, E.coli swims in a straight line. (P66)... When the
motors are synchronized again, E.Coli takes off in a new direction that
has been shown to be unrelated to the former direction.
The trick is that when E.COli is swimming up a gradient of some good
chemical in its surroundings, or down a gradient of a bad one, tumbling
is delayed, while if the direction is toward the bad or away from the
good, the time to the next tumble is shortened. ... On average, it (or
rather a simulation based on this principle) makes its way toward the
source of the good substance or away from the bad one over hal;f as fast
as it could do if it could just turn and swim the right way.
Suddenly random reorganization didn't look as inefficient as I thought
it would have to be. The key idea is that what is change by a
reorganization is not a position, not the value of some behavioural
parameter, but the rate of change of position or value. And what was
sensed was not the degree of goodness or badness, but the first
derivative of goodness or badnessï¿½whether the result was getting better
or worse. ...
In E. Coli what is reorganized is its direction of swimming...But the
direction of swimming through space can be treated as analogous to the
direction of altering a set of control system parameters in some
abstract multidimensional space, and a tumble can be trreated as an
alteration in the proportions by which the parameters are being changed,
so the direction of metaphorical swimming in this multidimensional space