The Blind Wathmaker

Download Report

Transcript The Blind Wathmaker

Game of Life
• Grid of binary elements, arranged on a 2-D
plane, like pixels on a computer screen
• Each cell belongs to a neighborhood including
its 8 immediate neighbors.
• Each cell follows simple local rules
• Cohesive patterns form out of apparent
randomness
• Emergence
Cellular automata
• Very simple virtual machine that results in
complex, even lifelike, behavior
• A cell is a bit-position on a binary string
• State of a cell at the next time step is
determined by its state at the current time
and by the states of cells in its neighborhood
• Simplest neighborhood includes a cell and its
neighbor on either side (a neighborhood of 3)
Dawkins’ biomorphs
• Richard Dawkins, The Blind Wathmaker, and
biomorphs
• Form was encoded in a chromosome of 9
genes, each could take values from 0 to 9.
• These encoded rules for the development of
the biomorph, e.g. angle or length of a branch
• Each gene can change on step by mutation
• User could generate a random population,
select one that looked interesting, nad use it
to generate a new population, and so forth
Evolution
• Dawkins found that, though he expected his
population to look like trees, he discovered
that insect forms were also possible
• Incremental and unpredictable change in a
desired direction: evolution
Accomplishments of Social Insects
• An insect may have only a few hundred brain
cells (compared to hundreds of trillions of
brain cells in a human), but insect
organizations are capable of architectural
marvels, elaborate communication systems,
and terrific resistance to threats of nature
• A central question in the sociobiology of
insects is: how does mass behavior emerge
from the behaviors of single ants
Termites
• Termites build domes by thru simple rules
• Autocatalytic or positive feedback cycle
• Termite builders are a self-organizing system,
there is no central control, the members of
the population are unaware of the “plan” they
are carrying out
• Stigmergy: communication by altering the
state of the environment in a way that will
affect the behavior of others for whom the
environment is a stimulus
Ants
•
•
•
•
Ant cemeteries
Formation of piles of woodchips by termites
Live ant experiment: ants find shorter path
If the length is changed during the
experiment, some species re-discover the
shorter path and some don’t
• Flies trapped in a jar
Fish Schools
• Many species of fish swim in schools that
seem to take on an emergent life of their own
• A fish school appears to move as one, with
hundreds if not hundreds of thousands
changing direction, darting at what appears to
be the same exact instant
• Experiment with a solitary fish in divided tank
• c=kNt (e.g. k=0.355, t=0.818)
• Effect of t<1: sublinear increase with groupsize
Social Impact
• Latane’s experiments on social impact find
that the impact of a group on an individual is a
function of the Strength, Immediacy, and
Number of sources of influence
• Nervousness of participants in a college talent
show shown to be a function of number of
people in audience (similar formula to fish)
• Do we humans exhibit herding behavior?
• Craig Reynolds bird flocking simulation
Fish Schooling
• Biologists have shown that fish seem to
regulate their schooling behavior using visual
information and information from the fish’s
lateral line
• Vision guides the flock-centering tendency
• Lateral line enables collision avoiding
• Blinded pollock swim farther from their
neighbors
• Fish with lateral lines removed swarm closer
to their neighbors than normal fish
Herds and birds
• Possible explanation for herd-centering
behavior: an animal at the edge of the herd is
more likely to be picked off by a predator
• Biologists carried out an experiment: bird
flocks were filmed with movie cameras filming
at 3 frames per second.
• Watching these one frame at a time, they
could distinguish specific features of flocking
• For example, there was no leader to the flock.
Any bird could lead a maneuver at any time
Shannon’s experiments
• First-order: each letter is selected
probabilistically based on its frequency
• Second-order: based on probability that letter
x will be followed by letter y
• Third-order: based on probability that a 2letter sequence is followed by another
• Experiments up to fifth order
• Second-order word approximation
Conformity and social norms
• Sherif placed subjects in a dark room with a
point of life projected (autokinetic effect)
• Asch experiments: subjects asked to rate
relative lengths of some lines (about a third
agreed with group)
• Subsequent experiments: subjects in booths,
so they could not see rest of group
Are humans herd animals?
• Economist Brian Arthur has noted that it is
impossible for people to reason deductively in
complex situations; there are just too many
linkages of facts for anyone to keep them
straight.
• People end up floundering in a pool of
subjective beliefs, including subjective beliefs
about subjective beliefs.
• The rationality that is assumed by classical
economists cannot hold
Drunk Irishmen
• El Farol is a bar that has Irish music on
Thursday nights
• Each irishman wants to go when it is not
crowded; prefers not to go if more than 60 are
there. Each would gladly go to El Farol, if only
he knew those other noisy bastards would
stay home
• One hundred agents set up, and each was
allowed to evolve a different strategy. Result:
attendance was nearly 60% over time
Drunk Irishmen Agents
• No agent dominated or performed any better
on average than any others
• Some very sophisticated deceptive strategies
had evolved
• Agents were allowed to report their intentions
and they were allowed to lie if they wished (of
course, if they lied too often, they lost
credibility)
Conformity
• Another experiment: students were asked to
choose one-side of a debate question to argue
for
• Student was told he was free to choose either
side, but was subtly persuaded to choose the
unpopular side (while still believing he freely
chose that side)
• Afterward the assignment was over, the
student was found to have incorporated some
of the beliefs of the side he had argued
Adaptive Culture Model
• Axelrod: “Agents who are similar to each other
are likely to interact and become more similar
• Experiment: each individual is represented as
a string of numerals (one might be 42237,
another 99217)
• A matrix of individuals is initialized
• An individual and one of its four neighbors are
selected at random
Adaptive Culture Model
• Based on flipping a biased coin (biased by
degree of intersection), they interact.
• One individual copies one numeral from the
other
• Sociologists say this is similar to humans, we
pick up habits from those who are similar to
us
• But, perhaps in humans, we pick up habits
from those who are similar to what we would
like to be (e.g. success of Seven Habits)
Adaptive Culture Model
• Experiment: optimize a simple function
• Experiment: Sum of first three digits equal to
sum of last two
• Requires complex coordination of entire
vector of elements
• Experiment: 8-city TSP.
• Global optimal found in 11 of 20 trials
Cultural Adaptation
• Process of cultural adaptation, at the lowestlevel, includes three principles:
– Evaluate your neighbors
– Compare to yourself
– Imitate your betters
Swarm Intelligence
• A system in which many individual agents with
limited intelligence and information are able
to pool resource to accomplish a goal beyond
the capabilities of the individuals.
Particle Swarm
•
•
•
•
Binary PSO
Real-valued PSO
Train a neural network or a fuzzy rule system
No Free Lunch Theorem