Preface
Intelligence “arises from interactions among individuals.”
pg xiii
Heuristics – “shortcuts to speed up the process of sorting through possibilities.”
“Programs that search through a geographical map can be easily adapted to explore deductive threads in other domains.” – particle swarm optimizations among them.
pg xiv
“Agent subroutines may pass information back and forth, but subroutines are not changed as a result of the interaction, as people are. Individuals exchange rules, tips, and beliefs about how to process the information. Thus social interaction typically results in a change in the thinking process – not just the contents – of the participants.”
“Schools of fish have an advantage in escaping predators, as each individual fish can be a kind of lookout for the whole group. Herding animals also have an advantage in finding food: if one animal finds something to eat, the others will watch and follow.”
“Intelligence is the ability to adapt.”
pg xv
Swarm “refers to a disorganized cluster of moving things, usually insects, moving irregularly, chaotically, somehow staying together even while all of them move in apparent random directions.”
Swarms have applications in “understand intelligence and aspects of culture”, and optimizations.
pg xvi
Evolutionary computation paradigms
(Eberhart, Simpson, and Dobbins, 1996)
- Genetic algorithms
- Evolutionary programming
- Evolution strategies
- Genetic programming
Particle swarm optimization
-“utilizes a ‘population’ of candidate solutions to evolve to an optimal or near-optimal solution to a problem.
-“The degree of optimality is measured by a fitness function defined by the user.”
-“differs from evolutionary computation methods in the the population members, called particles, are flown through the problem hyperspace.”
-initialized with random values and “stochastically assigned velocities.”
-“each iteration, each particle’s velocity is stochastically accelerated toward its previous best position (where it had its highest fitness value) and toward a neighborhood best position (the position of highest fitness by any particle in its neighborhood).”
-“powerful, easy to understand, easy to implement, and computationally intensive.”
-“central algorithm comprises just two lines of computer code”
-“often at least an order of magnitude faster than other evolutionary algorithms on benchmark functions.”
-“extremely resistant to being trapped in local optima.”
-“applied to fields as diverse as electric/hybrid vehicle battery pack state of charge, human performance assessment, and human tremor diagnosis.”
-“provides evidence for theoretical perspectives on mind, consciousness, and intelligence.” – intelligence is at least partly attributed to the ability to interact with others and base our behaviors on those interactions.
Cellular Automata (CA)
-“used for self-generated computer graphics movies, simulating biological systems and physical phenomena, designing massively parallel computers, and most importantly for basic research into the characteristics of complex dynamic systems.
-3 main attributes
-(1) individual cell updates are done in parallel, (2) each new cell value depends only on the old values of the cell and its neighbors, and (3) all cells are updated using the same rules. (Rucker, 1999).
pg xvii
Emergent Behavior
-“simple processes leading to complex results.”
-“the whole being more than the sum of its parts.”
-“There is no conservation of simplicity.” – Stephan Wolfram
xviii
Swarm Intelligence: From Natural to Artificial Systems by Bonabeau, Dorigo, and Theraulaz (1999)
-“very few applications of swarm intelligence.”
-“hard to ‘program’ because the paths to problem solving are not predifined but emergent in these systems and result from interactions among indivuals and between indiviudals and their environment as much as from the behaviors of the individuals themselves.”
-“requires a thorough knowledge not only of what individual behaviors must be implemented but also of what interactions are needed to produce such or such global behavior.”
-Swarm Intelligence – “the emergent collective intelligence of groups of simple agents.”
“quite a few applications of swarm intelligence have been developed.”
“easy to program”
“knowledge of individual behaviors and interactions is not needed. Rather, these behaviors and interactions emerge from very simple rules.”
“Members of a swarm seem to us to fall short of the usual qualifications for something to be called an ‘agent,’ notably autonomy and specialization. Swarm members tend to be homogeneous and follow their program explicitly.
Swarm vs flock vs school vs herd
-first programs modeled flocks and schools.
-“as the programs evolved from modeling social behaviors to doing optimization, the two-dimensional plots we used to watch the algorithms ceased to look much like bird flocks or fish schools and started looking more like swarms of mosquitoes.”
pg xix
5 Principles of Swarm Intelligence
by Mark Millonnas (1994)
- The proximity principle: The population should be able to carry out simple space and time computations.
- The quality principle: The population should be able to respond to quality factors in the environment.
- The principle of stability: The population should not change its mode of behavior every time the environment changes.
- The principle of adaptability: The population must be able to change behavior when it’s worth the computational price.
Particle
-“massless and volumeless mathematical abstractions and would be called ‘points’ if they stayed still;”
-“velocities and accelerations are more appropriately applied to particles, even if each is defined to have arbitrarily small mass and volume.
Particle Systems (Reeves, 1983)
-“clouds of primitive particles as models of diffuse objects such as clouds, fire, and smoke within a computer graphics framework.”
page xx
Assertions and corollaries
- Mind is social
- Human intelligence results from social interaction.
- Culture and cognition are inseparable consequences of human sociality.
- Particle swarms are a useful computational intelligence (soft computing) methodology.
- Swarm intelligence provides a useful paradigm for implementing adaptive systems.
- Particle swarm optimumization is an extension of, and potentially important new incarnation of, cellular automata.
pg xxi
Chapter One
Christopher Langton – developed Swarm with colleagues at Santa Fe Institute and the Swarm Corporation
Swarm – “any loosely structured collection of agents that interact with one another. In common usage, the term agent refers to an object in a computer program that has an identity and performs some actions, usually with some presumption that activities of the agent are somewhat autonomous or independent of the activities of the other agents. An ant colony is a kind of swarm where the agents are ants, highway traffic can be conceptualized as a swarm where the agents are cars and drivers.”
Hierarchical Swarms
“interactions between individual- and group-level phenomena”
“bird flock has properties over and above the properties of the birds themselves, though there is of course a direct relationship between the two levels of the phenomena.”
“An economy might be made up of a swarm of agents who are people, while each person might be made up of a swarm that is their beliefs or even the neurons that provide the infrastructure for their beliefs – swarms within swarms.”
pg 25
Flocks, Herds, Schools, and Swarms: Social Behavior as Optimization
Stochasticism as optimization technique
Mutation, reproduction, selection, and self-organization drive the process of optimization in evolution. Similar principles can apply to creative thinking and problem solving.
“Pyschologist Donal Campbell (1960) describes creative thinking in terms of ‘blind variation and selective retention.”
Random activity is useful for predator avoidance, searching for food/mate etc.”
Marine snail moves randomly from side to side in a seemingly random escape pattern as it tries to find the source of food via optimizing the strength of the scent on its receptor. – Lorenz (1973)
pg 94
Sociality as a selfish trait
Communal rearing
Sharing food, warmth, warning, defense.
“Groups of prey may be more conspicuous than solitary individuals but running with a pack reduces the probability of any individual being selected for attack.”
More eyes leads to greater foraging efficiency
“Noting where a well-fed flockmate just flew in from”
Predators beneft from coordinated hunting
Insects benefit from division of labor – building elaborate nests
pg 96
E. coli bacteria show ‘intelligence’ through adaptive behavior. This “adaptation is the cornerstone of the iterated function system model of neuronal interactions in the human brain.”- Hoskins (1995)
“Ben-Jacobs asserts that the lowly bacteria colony is capable of computing better than the best computers of our time, and attributes them to properties of creativity, intelligence, and even self-awareness.”
pg 97
Slime mold exist as single-celled amoeba when food is plentiful but bond together to form a slug-like form that can crawl when food is scarce.
Fixed action patterns – Lorenz
Imprinting, territorial marking have been shown to be behaviors that an “organism emits in response to a particular stimulus”
Ants and other insects operate through a combination of fixed action patterns.
Pheromones – “chemicals that possess a kind of odor that can be detected by other ants.”
“Pheromones comprise a medium for communication among the ants, allowing fixed action collaboration, the result of which is a group behavior that is adaptive where the individuals’ behaviors are not.”
“Given that the behavior of a single ant is almost random, with a stochastic tendency to gravitate toward paths that have been trodden by other ants, the achievement of swarms of ants are most incredible.”
pg 99
“Insect sociability is a classic example of the emergence of global effects from local interactions.”
much more than the sum of its parts
pg 100
a solitary ant simple and stupid and will perish soon. “ants succeed by collaborating and so do people”
1979 Douglas Hofstadter – brain is like ant colony. no single neuron contains knowledge, thinking can only occur through interactions.
Mitchel Resnick, 1998 – “ants have become the unofficial mascots of the ALife community.
Jean-Louis Deneubourg – model of ants can solve some kinds of combinatorial problems that were previously thought too hard to attempt
pg 101
“We use the word ‘swarm’ in a general sense to refer to any such loosely structured collection of interacting agents The classic example of a swam is a swarm of bees, but the metaphor of a swarm can be extended to other systems with a similar architecture. Ant colony can be thought of as a swarm whose individual agents are ants, a flock of birds is a swarm whose agents are birds, traffic is a swarm of cars, a crowd is a swarm of people, an immune system is a swarm of cells and molecules, and an economy is a swarm of economic agents. Although the notion of a swarm suggests an aspect of collective motion in space, as in the swarm of a flock of birds, we are interested in all types of collective behavior, not just spatial motion.” – Santa Fe Institute
pg 102
Termites are able to build elaborate domes. They begin constructing pillars and slowly tilt them towards each other until the tops touch and thus end up with an arch. By connecting multiple arches, the termites create a dome, a structure that even eluded man for thousands of years and whose creation marked a major milestone in architecture. The approach man takes in building a dome is a top down one: start with a plan and have a leader coordinate every team member and brick placement in an ordered succession. Termites manage to create the same structure with a distributed, decentralized methodology that is incredibly scalable and able to be built with a very low level rule set: 1) Pick up dirt in your mouth to moisten it 2) Move in the direction of the strongest pheromone concentration and 3) Deposit what you are carrying where the smell is the strongest. Such simple rules lead to a self-organizing system where the workers do not even have to be aware of the ‘plan,’ the structure just emerges without symbolic communication. Architects have long been and should be envious of termites for finding ways to build such improbable constructs without dealing with the problem of communication issues. Just imagine the possibilities if we are able to harness the same processes that drive organic systems.
pg 103-104
Stigmergy – P.P. Grasse 1959 – a method of indirect communication though the altering of the state of the environment in such way that it will stimulate a behavior in others who encounter the altered state.
Cue-based stigmergy – pillars for termites, ant corpse piling, woodchip gathering in termites,
Sign-based stigmergy – pheromones
pg 104-105
Optimizing with Simulated Ants
Dorigo and Gambardella 1997 “showed how very simple pheromone-following behavior could be used to optimize the traveling salesman problem. Their ‘ant colony optimization’ is based on the observation that ants will find the shortest path around an obstacle separating their est from a target such as a piece of candy”
Ants leave pheromones as they walk that dissapate over time and thus if a spot is more traveled or traveled recently, it will have a stronger pheromone scent. Thus, shorter paths will have stronger pheromone trails. This algorithm can solve problems where the goal is to “accomplish a task in the fewest number of operations.”
pg 106
Staying Together but Not Colliding: Flocks, Herds, and Schools
Breder 1954 showed that fish are attracted to groups, but the impact of an additional fish was much greater when the group size was small than when the group was much larger. So, as the number of fish increases, the effect on each fish increases, but the rate of increase slows.
pg 109-110
Latane 1981 shows that this effect is similar on humans as shows through experiments with group tipping at restaurants and the number of people in the audience at a talent show’s effect on participant’s nervousness. Social psychology has shows that humans are flocking, schooling, herding creatures not in the physical way we move, except for crowd movement, but more so in the way we influence each other’s thoughts and actions. It has been shown that when people interact, they become more similar. This is the medium through which culture is born.
p 110-111
Craig Reynolds 1987 – “flocking birds were driven by three local forces: collision avoidance, velocity matching, and flock centering. That is ‘boids’ (Reynolds’ name for simulated birds)
- pull away before they crash into one another
- try to go about the same speed as their neighbors in the flock
try to move toward the center of the flock as they perceive it
”
pg 111
W.D. Hamilton 1971 – ‘Geometry of the selfish herd’ theorized why animals flock, herd, and school. He pointed out that these behaviors are much more often seen in prey animals than predators. Prey have a great incentive to be in the center of their population as those on the outside are in a much more dangerous position of being picked off. Thus the cohesion behavior of prey animals that Reynolds simulated is actually a ‘selfish’ act by each population member. The other two behaviors, alignment and separation, are equally selfish as it is in no one’s best interest to crash into each other. By alignment velocities and turning the other way when a neighbors gets too close, these behaviors act as a type of collision avoidance technique.
p 112 – 113
Reynolds 1987 compared his model of flocking to a particle system.
Frank Heppner, a biologist, was conducting similar but more scientific research into bird flocks around the same time as Reynolds. He concluded that there was no ‘leader’ in a flock; any flock member could lead a movement at any time and the rest would update their velocities. His other findings of an attractive/repulsive force and velocity matching lined up perfectly with Reynolds theory.
pg 113
The way birds find food on the ground while flying high in the air is an interesting case study in another social behavior: seeing someone see something. For instance, rather than looking for the food itself, which may be a small seed, birds look for cues that food is nearby. One such cue would be observing other birds eating or circling an area. Another could be seeing a neighbor turn back and descending toward a potential food source.
The function optimum of the particle swarm algorithm is called a ‘cornfield vector.’ The concept is based on finding a hidden point in a landscape using clues from other searchers.
pg 114
Agency
Agents seem “to have something like a will, or an ability to make judgements or decisions; it’s able to do things – and nit just what it’s told, either.”
Franklin and Graesser, 1996 – “An autonomous agent is a system situated within and a part of an environment that senses that environment and acts on it, over time, in pursuit of its own agenda and so as to effect what it senses in the future.”
We tend to anthropomorphize agents by ascribing things such as ‘sensations’, ‘actions’, ‘understanding’ and even ‘agendas’ to them.
p 129
“Picking parasites from one another’s fur us a very important social activity in most species of primates, allowing them to show affection and respect. apologize for aggression and strengthen alliances. Though grooming confers some small survival advantage by getting rid of parasites, this advantage clearly does not justify the amount of time spent at it. The same can be said for human conversation: the amount of information conveyed in conversation rarely justifies the amount of time people spend talking with one another. The primary purpose of both grooming and language is to establish and maintain social relationships.”
p 131