cs621-lect7-SI-13aug07.ppt

Swarm Intelligence
From Natural to Artificial Systems
mohitz, bhavish, amitb, madhusudhan

Swarming – The Definition
 aggregation of similar animals, generally
cruising in the same direction
 Termites swarm to build colonies
 Birds swarm to find food
 Bees swarm to reproduce

Why do animals swarm?
 To forage better
 To migrate
 As a defense against predators
 Social Insects have survived for millions of
years.

Swarming is Powerful
 Swarms can achieve things that an individual
cannot

Swarming – Example
 Bird Flocking
 “Boids” model was proposed by Reynolds
 Boids = Bird-oids (bird like)
 Only three simple rules

Collision Avoidance
 Rule 1: Avoid Collision with neighboring birds

Velocity Matching
 Rule 2: Match the velocity of neighboring
birds

Flock Centering
 Rule 3: Stay near neighboring birds

Swarming - Characteristics
 Simple rules for each individual
 No central control
 Decentralized and hence robust
 Emergent
 Performs complex functions

Learn from insects
 Computer Systems are getting complicated
 Hard to have a master control
 Swarm intelligence systems are:
 Robust
 Relatively simple

Swarm Intelligence - Definition
 “any attempt to design algorithms or
distributed problem-solving devices inspired
by the collective behavior of social insect
colonies and other animal societies”
[Bonabeau, Dorigo, Theraulaz: Swarm
Intelligence]
 Solves optimization problems

Applications
 Movie effects
 Lord of the Rings
 Network Routing
 ACO Routing
 Swarm Robotics
 Swarm bots

Roadmap
 Particle Swarm Optimization
 Applications
 Algorithm
 Ant Colony Optimization
 Biological Inspiration
 Generic ACO and variations
 Application in Routing
 Limitations of SI
 Conclusion

Particle Swarm Optimization
Madhu Sudhan

Particle Swarm Optimization
 Particle swarm optimization imitates human
or insects social behavior.
 Individuals interact with one another while
learning from their own experience, and
gradually move towards the goal.
 It is easily implemented and has proven both
very effective and quick when applied to a
diverse set of optimization problems.

 Bird flocking is one of the best example of
PSO in nature.
 One motive of the development of PSO was
to model human social behavior.

Applications of PSO
 Neural networks like Human tumor analysis,
Computer numerically controlled milling
optimization;
 Ingredient mix optimization;
 Pressure vessel (design a container of
compressed air, with many constraints).
Basically all the above applications fall in a
category of finding the global maxima of a
continuous, discrete, or mixed search space,
with multiple local maxima.

Algorithm of PSO
 Each particle (or agent) evaluates the
function to maximize at each point it visits in
spaces.
 Each agent remembers the best value of the
function found so far by it (pbest) and its co-
ordinates.
 Secondly, each agent know the globally best
position that one member of the flock had
found, and its value (gbest).

Algorithm – Phase 1 (1D)
 Using the co-ordinates of pbest and gbest,
each agent calculates its new velocity as:
vi = vi + c1 x rand() x (pbestxi – presentxi)
+ c2 x rand() x (gbestx – presentxi)
where 0 < rand() <1
presentxi = presentxi + (vi x Δt)

Algorithm – Phase 2 (n-dimensions)
 In n-dimensional space :

Ant Colony Optimization
Bhavish

Ant Colony Optimization - Biological
Inspiration
 Inspired by foraging behavior of ants.
 Ants find shortest path to food source from nest.
 Ants deposit pheromone along traveled path which
is used by other ants to follow the trail.
 This kind of indirect communication via the local
environment is called stigmergy.
 Has adaptability, robustness and redundancy.

Foraging behavior of Ants
 2 ants start with equal probability of going on
either path.

 The ant on shorter path has a shorter to-and-
fro time from it’s nest to the food.

 The density of pheromone on the shorter
path is higher because of 2 passes by the ant
(as compared to 1 by the other).

 The next ant takes the shorter route.

 Over many iterations, more ants begin using
the path with higher pheromone, thereby
further reinforcing it.

 After some time, the shorter path is almost
exclusively used.

Generic ACO
 Formalized into a metaheuristic.
 Artificial ants build solutions to an
optimization problem and exchange info on
their quality vis-à-vis real ants.
 A combinatorial optimization problem reduced
to a construction graph.
 Ants build partial solutions in each iteration
and deposit pheromone on each vertex.

Ant Colony Metaheuristic
 ConstructAntSolutions: Partial solution extended by adding
an edge based on stochastic and pheromone
considerations.
 ApplyLocalSearch: problem-specific, used in state-of-art
ACO algorithms.
 UpdatePheromones: increase pheromone of good
solutions, decrease that of bad solutions (pheromone
evaporation).

Various Algorithms
 First in early 90’s.
 Ant System (AS):
 First ACO algorithm.
 Pheromone updated by all ants in the iteration.
 Ants select next vertex by a stochastic function
which depends on both pheromone and problem-
specific heuristic nij = 1
dij

Various Algorithms - 2
 MAX-MIN Ant System (MMAS):
 Improves over AS.
 Only best ant updates pheromone.
 Value of pheromone is bound.
 Lbest is length of tour of best ant.
 Bounds on pheromone are problem specific.

Various Algorithms - 3
 Ant Colony System (ACS):
 Local pheromone update in addition to offline
pheromone update.
 By all ants after each construction step only to last
edge traversed.
 Diversify search by subsequent ants and produce
different solutions in an iteration.
 Local update:
 Offline update:

Theoretical Details
 Convergence to optimal solutions has been
proved.
 Can’t predict how quickly optimal results will
be found.
 Suffer from stagnation and selection bias.

ACO in Network Routing
Amit Bharadwaj

Ant like agents for routing
 Intuitive to think of ants for routing problem
 Aim is to get shortest path
 Start as usual
 Release a number of ants from source, let the age
of ant increases with increase in hops
 decide on pheromone trails i.e biasing the entries in
routing table in favor of youngest ant
 Problem – Ants at an node do not know the
path to destiation, can't cahnge table entry

Routing continued ...
 Possible Solutions
 first get to dest. and then retrace
 Needs memory to store the path
 And intelligence to revert the path
 Leave unique entries on nodes
 a lot of entries at every node
 Observation – At any intermediate node, ant
knows the path to source from that node.
 now leave influence on routing table having entry
“route to source via that link”

Routing contd ...
 Now at any node it has information about
shortest path to dest., left by ants from dest.
 The ant following shortest path should have
maximum influence
 A convenient form of pheromone can be
inverse of age + constant
 The table may get frozen, with one entry
almost 1, add some noise f i.e probabilty that
an ant choses purely random path

Dealing with congestion
 Add a function of degree of congestion of each
node to age of an ant
 Delay an ant at congested node, this prevents
ants from influencing route table

SI - Limitations
 Theoretical analysis is difficult, due to
sequences of probabilistic choices
 Most of the research are experimental
 Though convergence in guaranteed, time to
convergence is uncertain

Scope
 Startup !!
 Bluetronics, Smartintel
 Analytic proof and models of swarm based
algorithm remain topics of ongoing research
 List of applications using SI growing fast
 Controlling unmanned vehicles.
 Satellite Image Classification
 Movie effects

Conclusion
 Provide heuristic to solve difficult problems
 Has been applied to wide variety of
applications
 Can be used in dynamic applications

References
 Reynolds, C. W. (1987) Flocks, Herds, and Schools: A Distributed Behavioral
Model, in Computer Graphics, 21(4) (SIGGRAPH '87 Conference
Proceedings) pages 25-34.
 James Kennedy, Russell Eberhart. Particle Swarm Optimization, IEEE Conf.
on Neural networks – 1995
 www.adaptiveview.com/articles/ ipsop1
 M.Dorigo, M.Birattari, T.Stutzle, Ant colony optimization – Artificial Ants as a
computational intelligence technique, IEEE Computational Intelligence
Magazine 2006
 Ruud Schoonderwoerd, Owen Holland, Janet Bruten - 1996. Ant like agents
for load balancing in telecommunication networks, Adaptive behavior, 5(2).

cs621-lect7-SI-13aug07.ppt

More Related Content

Similar to cs621-lect7-SI-13aug07.ppt (20)

Recently uploaded (20)

cs621-lect7-SI-13aug07.ppt