Exploring Usage Of Meta-heuristics in Engineering

1. Ossein Jain
Mtech (CS)
By-

2. What is ?
In simple words:
• It is a repetitive process used to solve a
problem that even the toughest algorithm,
with greatest optimization powers, fail to be
effective or efficient.

4. Artificial Neural Networks
• Most Basic Technique of AI-“ANN (Artificial neural
Networks)” , based on Human nervous system
• Processes information with the help of
– nodes
– synapses(inputs)
– weights(signals)
– mathematical formulas (calculating neuron output signal,
scalar product i.e. net ).
•Neuron Communicating through
input, hidden and output layers
and weighted connections
(feedforward, feedback, lateral and
time-delayed).

5. ANN continued....
• An ANN problems is solved through
• Training
• Generalization (Testing)
• Implementation.
• ANN is applied in Robotics, Data Processing, Functional
approximation and pattern sequence recognition.
• Disadvantage is that because the network finds out how to solve
the problem by itself and its operation can be unpredictable

6. Tabu Search (TS)
• Taking a solution as starting point (or local minima), uses a
Tabu list for recent moves and a Tabu memory to prevent it
from repeating, making a “Solution space - Diversification”.
When we need better version of better
Initialize
solution
Populate
candidate
list of
solutions
Evaluate
solutions
Choose
best
admissible
solution
Stopping
conditions
satisfied?
Update
aspiration
conditions
Update
tabu Final
soution
yes
no

7. Tabu search
• Many factors play a role like size, adaptability of TS memory and
list, local search procedure, form of moves and stopping rule.

8. Evolutionary Algorithm
• When we have a bundle of solutions and we know not to give up, we
use Evolutionary Algorithm
• Methods like Genetic programming, Evolutionary strategies, Genetic
Algorithms are used.
• Unlike TS it has many “local optimas”.
• EAs are used in like wire routing, scheduling, image processing, game
playing, Knapsack problem, etc.
mixing and matching the best part of each solution

9. Genetic Algorithm
• GA is an adaptive search and optimization using
random searches to find “local optimal
solutions”, so as to safeguard some critical info.
•Mutate, then select
the fittest solution and
repeat until the best
one is found.
• Popular for
Bankruptcy prediction,
residual estimation,
vehicle routing, etc.

10. Differential Evolution
• It makes a trial vectors using existing solutions and
mixes it with successful ones, further improved by
mutation, crossover and selection operators.
• DE Algorithm:
Advanced Version of GA that focuses on Mutation
• Best for numerical problems, used to find
approximate solutions where problems are non
linear, non-differentiable with many local minima and
constraints

11. Simulated Annealing
• A worse variation is accepted as the new solution with a
probability that decreases as the computation proceeds.
• The slower the cooling schedule, or rate of decrease, the
more likely the algorithm is to find an optimal or near-
optimal solution
• It is Useful in zoning, routing, facility layout problems.
When in need to find random variations of a present
solution, accepting the worst one

12. Swarm Intelligence
• Swarm Intelligence (SI), follows 5 principles:
proximity, quality, diverse response, stability
and adaptability.
Inspired from insects and their coordinated interactive
teamwork

13. Particle Swarm Optimization (PSO)
• Particle Swarm Optimization (PSO) is based
on population on concept of bird flocking.
• It is easy parallelization for concurrent
processing, derivative free and solve
convergence is very effective.
• Implemented in Parkinson’s disease
identification, electric power distribution,
biometrics, processing biochemistry, etc

14. Ant colony Optimization (ACO)
• Ant colony Optimization (ACO) is inspired by foraging and
colonization of ants.
• It includes trailing like ants, making progressive solutions,
using attractiveness and trail levels.
• ACO algorithms like Ant system, Ant colony system, ma-
min ant system, rank based ant system and best-worst
ant system are summarised.
• In Ant System, the contributions by ants depends
on quality of solution and better the trail contribution,
better the solution.

15. Ant colony Optimization (ACO)…
• To improve the algorithmic quality, performance and
behaviour ACS, enhanced AS through pseudo-random
proportion rule, updating pheromone trail offline through
daemon, hence not every ant follows the same ant.
• MMASas best enhancement of AS, ranking ant in decreasing
order of respective solutions.
• Pheromone which restricts them was influenced by rand and
quality of solution, so was the connection.
• BWAS using the transition rule and Pheromone
evaporation techniques, was also a good extension to
AS.
• ACO was largely useful in assignment, scheduling,
vehicle routing, travelling salesman and energy
forecasting

16. FUTURE RESEARCH DIRECTIONS
• However useful these techniques were, they did not guarantee an
optimal solution, overhead being , complex function , parameters and
constraints, also lack of standards of testing and comparison of
methods makes it rather in need for improvement.
• As in latest 5 years, Metaheuristic have gained importance through
textbooks, conferences, success in application many real-world
problems.
• ANNs and GAs softwares being available are more popular.
• As the better optimal solution are much needed for the scarcity of
time, money and resources, so will just keep on increasing in near
future .Thus the number of methods for Metaheuristic.
• All we need to do is, develop softwares for all such useful methods to
be better applied in the real-world.