Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• A search technique used in computing to find near optimal
solutions to discrete optimization problems.
• ACO is a swarm intelligence inspired from the way that ants
indirectly communicate directions to each other.
• The most interesting aspect of the collaborative behaviour of ant
species is their ability to find the shortest paths between the ants'
nest and the food sources.
• Finding the shortest path between their nest and a food source by
chase pheromone trails exposed by other ants. The more intense the
trail, the higher probability that an ant will follow it and thus enrich
the trail with its own pheromone. The pheromone trails of longest
paths evaporate.
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• A search technique used in computing to find near optimal
solutions to discrete optimization problems.
• ACO is a swarm intelligence inspired from the way that ants
indirectly communicate directions to each other.
• The most interesting aspect of the collaborative behaviour of ant
species is their ability to find the shortest paths between the ants'
nest and the food sources.
• Finding the shortest path between their nest and a food source by
chase pheromone trails exposed by other ants. The more intense the
trail, the higher probability that an ant will follow it and thus enrich
the trail with its own pheromone. The pheromone trails of longest
paths evaporate.
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• A search technique used in computing to find near optimal
solutions to discrete optimization problems.
• ACO is a swarm intelligence inspired from the way that ants
indirectly communicate directions to each other.
• The most interesting aspect of the collaborative behaviour of ant
species is their ability to find the shortest paths between the ants'
nest and the food sources.
• Finding the shortest path between their nest and a food source by
chase pheromone trails exposed by other ants. The more intense the
trail, the higher probability that an ant will follow it and thus enrich
the trail with its own pheromone. The pheromone trails of longest
paths evaporate.
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• A search technique used in computing to find near optimal
solutions to discrete optimization problems.
• ACO is a swarm intelligence inspired from the way that ants
indirectly communicate directions to each other.
• The most interesting aspect of the collaborative behaviour of ant
species is their ability to find the shortest paths between the ants'
nest and the food sources.
• Finding the shortest path between their nest and a food source by
chase pheromone trails exposed by other ants. The more intense the
trail, the higher probability that an ant will follow it and thus enrich
the trail with its own pheromone. The pheromone trails of longest
paths evaporate.
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• A search technique used in computing to find near optimal
solutions to discrete optimization problems.
• ACO is a swarm intelligence inspired from the way that ants
indirectly communicate directions to each other.
• The most interesting aspect of the collaborative behaviour of ant
species is their ability to find the shortest paths between the ants'
nest and the food sources.
• Finding the shortest path between their nest and a food source by
chase pheromone trails exposed by other ants. The more intense the
trail, the higher probability that an ant will follow it and thus enrich
the trail with its own pheromone. The pheromone trails of longest
paths evaporate.
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Key Terms
 Ants 𝑘: Any possible solution.
 Population 𝑁- Group of all ants.
 Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.
 Search Space is divided by step size ℎ
 Pheromone trail 𝜏
 Scaling parameter 𝜁
 Evaporate rate ρ
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Key Terms
 Ants 𝑘: Any possible solution.
 Population 𝑁- Group of all ants.
 Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.
 Search Space is divided by step size ℎ
 Pheromone trail 𝜏
 Scaling parameter 𝜁
 Evaporate rate ρ
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Key Terms
 Ants 𝑘: Any possible solution.
 Population 𝑁- Group of all ants.
 Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.
 Search Space is divided by step size ℎ
 Pheromone trail 𝜏
 Scaling parameter 𝜁
 Evaporate rate ρ
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Key Terms
 Ants 𝑘: Any possible solution.
 Population 𝑁- Group of all ants.
 Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.
 Search Space is divided by step size ℎ
 Pheromone trail 𝜏
 Scaling parameter 𝜁
 Evaporate rate ρ
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Key Terms
 Ants 𝑘: Any possible solution.
 Population 𝑁- Group of all ants.
 Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.
 Search Space is divided by step size ℎ
 Pheromone trail 𝜏
 Scaling parameter 𝜁
 Evaporate rate ρ
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Key Terms
 Ants 𝑘: Any possible solution.
 Population 𝑁- Group of all ants.
 Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.
 Search Space is divided by step size ℎ
 Pheromone trail 𝜏
 Scaling parameter 𝜁
 Evaporate rate ρ
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Key Terms
 Ants 𝑘: Any possible solution.
 Population 𝑁- Group of all ants.
 Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.
 Search Space is divided by step size ℎ
 Pheromone trail 𝜏
 Scaling parameter 𝜁
 Evaporate rate ρ
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Key Terms
 Ants 𝑘: Any possible solution.
 Population 𝑁- Group of all ants.
 Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.
 Search Space is divided by step size ℎ
 Pheromone trail 𝜏
 Scaling parameter 𝜁
 Evaporate rate ρ
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Initialization
 Assume a suitable number of ants in the colony (population 𝑁)
 Assume a set of permissible discrete values 𝑚 for each of the design variables (step size ℎ).
 Initialize all discrete values of design variables equal amounts of pheromone 𝜏.
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Initialization
 Assume a suitable number of ants in the colony (population 𝑁)
 Assume a set of permissible discrete values 𝑚 for each of the design variables (step size ℎ).
 Initialize all discrete values of design variables equal amounts of pheromone 𝜏.
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Initialization
 Assume a suitable number of ants in the colony (population 𝑁)
 Assume a set of permissible discrete values 𝑚 for each of the design variables (step size ℎ).
 Initialize all discrete values of design variables equal amounts of pheromone 𝜏.
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Initialization
 Assume a suitable number of ants in the colony (population 𝑁)
 Assume a set of permissible discrete values 𝑚 for each of the design variables (step size ℎ).
 Initialize all discrete values of design variables equal amounts of pheromone 𝜏.
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Build tours
 From the home node, ants start travelling through the various paths and end
at the destination node in each iteration (discrete values of design variables).
 The probability to select discrete values of design variables is pj
k
=
τj
τj
m
j=1
 Find the cumulative probability ranges associated with different discrete
values based on its probabilities.
 The specific discrete values chosen by ant k will be determined using the
roulette-wheel selection.
Roulette-wheel
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Build tours
 From the home node, ants start travelling through the various paths and end
at the destination node in each iteration (discrete values of design variables).
 The probability to select discrete values of design variables is pj
k
=
τj
τj
m
j=1
 Find the cumulative probability ranges associated with different discrete
values based on its probabilities.
 The specific discrete values chosen by ant k will be determined using the
roulette-wheel selection.
Roulette-wheel
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Build tours
 From the home node, ants start travelling through the various paths and end
at the destination node in each iteration (discrete values of design variables).
 The probability to select discrete values of design variables is pj
k
=
τj
τj
m
j=1
 Find the cumulative probability ranges associated with different discrete
values based on its probabilities.
 The specific discrete values chosen by ant k will be determined using the
roulette-wheel selection.
Roulette-wheel
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Build tours
 From the home node, ants start travelling through the various paths and end
at the destination node in each iteration (discrete values of design variables).
 The probability to select discrete values of design variables is pj
k
=
τj
τj
m
j=1
 Find the cumulative probability ranges associated with different discrete
values based on its probabilities.
 The specific discrete values chosen by ant k will be determined using the
roulette-wheel selection.
Roulette-wheel
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Build tours
 From the home node, ants start travelling through the various paths and end
at the destination node in each iteration (discrete values of design variables).
 The probability to select discrete values of design variables is pj
k
=
τj
τj
m
j=1
 Find the cumulative probability ranges associated with different discrete
values based on its probabilities.
 The specific discrete values chosen by ant k will be determined using the
roulette-wheel selection.
Roulette-wheel
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
2
7
4
11
9
6
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
design variables
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
2
design variables
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
2
7
design variables
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
2
7
4
11
9
6
design variables
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
2
7
4
11
9
6
design variables
cumulative probability for
discrete values of design
variables
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
2
7
4
11
9
6
design variables
cumulative probability for
discrete values of design
variables
0.1
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
2
7
4
11
9
6
design variables
cumulative probability for
discrete values of design
variables
0.1
0.3
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
2
7
4
11
9
6
design variables
cumulative probability for
discrete values of design
variables
0.1
0.3
0.1
0.1
0.3
0.1
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
2
7
4
11
9
6
design variables
cumulative probability for
discrete values of design
variables
0.1
0.3
0.1
0.1
0.3
0.1
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
2
7
4
11
9
6
design variables
cumulative probability for
discrete values of design
variables
0.1
0.3
0.1
0.1
0.3
0.1
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
2
7
4
11
9
6
design variables
cumulative probability for
discrete values of design
variables
0.1
0.3
0.1
0.1
0.3
0.1
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
2
7
4
11
9
6
design variables
cumulative probability for
discrete values of design
variables
0.1
0.3
0.1
0.1
0.3
0.1
0.48
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Select Path (solution)
 Generate N random numbers r in the range (0, 1), one for each ant.
 Determine the discrete value by ant k for variable as the one for which the
cumulative probability range includes the random numbers r.
𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4
Roulette-wheel
2
7
𝟒
11
9
6
design variables
cumulative probability for
discrete values of design
variables
0.1
0.3
0.1
0.1
0.3
0.1
0.48
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Deposit and update trail
 Evaluate the objective function values of each ant
 Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants
 Update the pheromone
 Best ants: reinforcement the pheromone of the best path by: 𝜏𝑗
𝑛𝑒𝑤
← 𝜏𝑗
𝑜𝑙𝑑
+ ∆𝜏𝑗
𝑘
𝑘
 Other ants: evaporates the pheromone of other paths by: 𝜏𝑗
𝑛𝑒𝑤
← 1 − ρ 𝜏𝑗
𝑜𝑙𝑑
∆𝜏𝑗
𝑘
=
𝜁𝑓𝑏𝑒𝑠𝑡
𝑓𝑤𝑜𝑟𝑠𝑡
𝜁 → scaling parameter
𝑓𝑏𝑒𝑠𝑡 → best objective function
𝑓𝑤𝑜𝑟𝑠𝑡 → worst best objective function
ρ → evaporate rate
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Deposit and update trail
 Evaluate the objective function values of each ant
 Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants
 Update the pheromone
 Best ants: reinforcement the pheromone of the best path by: 𝜏𝑗
𝑛𝑒𝑤
← 𝜏𝑗
𝑜𝑙𝑑
+ ∆𝜏𝑗
𝑘
𝑘
 Other ants: evaporates the pheromone of other paths by: 𝜏𝑗
𝑛𝑒𝑤
← 1 − ρ 𝜏𝑗
𝑜𝑙𝑑
∆𝜏𝑗
𝑘
=
𝜁𝑓𝑏𝑒𝑠𝑡
𝑓𝑤𝑜𝑟𝑠𝑡
𝜁 → scaling parameter
𝑓𝑏𝑒𝑠𝑡 → best objective function
𝑓𝑤𝑜𝑟𝑠𝑡 → worst best objective function
ρ → evaporate rate
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Deposit and update trail
 Evaluate the objective function values of each ant
 Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants
 Update the pheromone
 Best ants: reinforcement the pheromone of the best path by: 𝜏𝑗
𝑛𝑒𝑤
← 𝜏𝑗
𝑜𝑙𝑑
+ ∆𝜏𝑗
𝑘
𝑘
 Other ants: evaporates the pheromone of other paths by: 𝜏𝑗
𝑛𝑒𝑤
← 1 − ρ 𝜏𝑗
𝑜𝑙𝑑
∆𝜏𝑗
𝑘
=
𝜁𝑓𝑏𝑒𝑠𝑡
𝑓𝑤𝑜𝑟𝑠𝑡
𝜁 → scaling parameter
𝑓𝑏𝑒𝑠𝑡 → best objective function
𝑓𝑤𝑜𝑟𝑠𝑡 → worst best objective function
ρ → evaporate rate
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Deposit and update trail
 Evaluate the objective function values of each ant
 Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants
 Update the pheromone
 Best ants: reinforcement the pheromone of the best path by: 𝜏𝑗
𝑛𝑒𝑤
← 𝜏𝑗
𝑜𝑙𝑑
+ ∆𝜏𝑗
𝑘
𝑘
 Other ants: evaporates the pheromone of other paths by: 𝜏𝑗
𝑛𝑒𝑤
← 1 − ρ 𝜏𝑗
𝑜𝑙𝑑
∆𝜏𝑗
𝑘
=
𝜁𝑓𝑏𝑒𝑠𝑡
𝑓𝑤𝑜𝑟𝑠𝑡
𝜁 → scaling parameter
𝑓𝑏𝑒𝑠𝑡 → best objective function
𝑓𝑤𝑜𝑟𝑠𝑡 → worst best objective function
ρ → evaporate rate
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Deposit and update trail
 Evaluate the objective function values of each ant
 Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants
 Update the pheromone
 Best ants: reinforcement the pheromone of the best path by: 𝜏𝑗
𝑛𝑒𝑤
← 𝜏𝑗
𝑜𝑙𝑑
+ ∆𝜏𝑗
𝑘
𝑘
 Other ants: evaporates the pheromone of other paths by: 𝜏𝑗
𝑛𝑒𝑤
← 1 − ρ 𝜏𝑗
𝑜𝑙𝑑
∆𝜏𝑗
𝑘
=
𝜁𝑓𝑏𝑒𝑠𝑡
𝑓𝑤𝑜𝑟𝑠𝑡
𝜁 → scaling parameter
𝑓𝑏𝑒𝑠𝑡 → best objective function
𝑓𝑤𝑜𝑟𝑠𝑡 → worst best objective function
ρ → evaporate rate
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Deposit and update trail
 Evaluate the objective function values of each ant
 Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants
 Update the pheromone
 Best ants: reinforcement the pheromone of the best path by: 𝜏𝑗
𝑛𝑒𝑤
← 𝜏𝑗
𝑜𝑙𝑑
+ ∆𝜏𝑗
𝑘
𝑘
 Other ants: evaporates the pheromone of other paths by: 𝜏𝑗
𝑛𝑒𝑤
← 1 − ρ 𝜏𝑗
𝑜𝑙𝑑
∆𝜏𝑗
𝑘
=
𝜁𝑓𝑏𝑒𝑠𝑡
𝑓𝑤𝑜𝑟𝑠𝑡
𝜁 → scaling parameter
𝑓𝑏𝑒𝑠𝑡 → best objective function
𝑓𝑤𝑜𝑟𝑠𝑡 → worst best objective function
ρ → evaporate rate
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Termination
 The steps of ACO algorithm are iteratively repeated until the maximum number of iteration
is reached or a termination criterion is met.
 Convergence: is the case where the positions of all particles converge to the same set of
values, the method is assumed to have converged.
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Termination
 The steps of ACO algorithm are iteratively repeated until the maximum number of iteration
is reached or a termination criterion is met.
 Convergence: is the case where the positions of all particles converge to the same set of
values, the method is assumed to have converged.
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Procedure
 Termination
 The steps of ACO algorithm are iteratively repeated until the maximum number of iteration
is reached or a termination criterion is met.
 Convergence: is the case where the of all ants converge to the same set of values, the
method is assumed to have converged.
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Parameters required from user:
 Population size 𝑁
 Set of permissible discrete values 𝑚 for each of the design variables
 Step size ℎ
 Initial pheromone trail 𝜏
 Scaling parameter 𝜁
 Evaporate rate ρ
 Termination criteria (i.e. number of iteration 𝑇)
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Pseudo code
1. Input
 Objective function (fitness function), upper bound (𝑢𝑏) and lower bound (𝑙𝑏), population size (𝑁), number of iteration 𝑇, scaling parameter 𝜁, evaporate rate ρ,
step size ℎ (or number of discrete value 𝑚)
2. Initialization
 Assume a set of permissible discrete values 𝑚 for each of the design variables
 Initialize all discrete values 𝑚 of design variables equal amounts of pheromone 𝜏
3. Loop:
 For 𝑡 = 1: 𝑇
 Find probability to select discrete values of design variables is 𝑝𝑗
𝑘
=
𝜏𝑗
𝜏𝑗
𝑚
𝑗=1
 Find the cumulative probability ranges associated with different discrete values based on its probabilities.(design roulette-wheel)
 For 𝑖 = 1: 𝑁 
 Generate a random numbers 𝑟, Find corresponding discrete value , Evaluate the objective function 𝑓𝑥𝑗
 Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants
 Update best path by: 𝜏𝑗
𝑛𝑒𝑤 ← 𝜏𝑗
𝑜𝑙𝑑 + ∆𝜏𝑗
𝑘
𝑘 and other paths by:𝜏𝑗
𝑛𝑒𝑤
← 1 − ρ 𝜏𝑗
𝑜𝑙𝑑
 If there is no convergence of the current solution & if 𝑡 > 𝑇 go to Loop
4. Print 𝑥𝑓𝑏𝑒𝑠𝑡 and 𝑓𝑏𝑒𝑠𝑡
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Pseudo code
1. Input
 Objective function (fitness function), upper bound (𝑢𝑏) and lower bound (𝑙𝑏), population size (𝑁), number of iteration 𝑇, scaling parameter 𝜁, evaporate rate ρ,
step size ℎ (or number of discrete value 𝑚)
2. Initialization
 Assume a set of permissible discrete values 𝑚 for each of the design variables
 Initialize all discrete values 𝑚 of design variables equal amounts of pheromone 𝜏
3. Loop:
 For 𝑡 = 1: 𝑇
 Find probability to select discrete values of design variables is 𝑝𝑗
𝑘
=
𝜏𝑗
𝜏𝑗
𝑚
𝑗=1
 Find the cumulative probability ranges associated with different discrete values based on its probabilities.(design roulette-wheel)
 For 𝑖 = 1: 𝑁 
 Generate a random numbers 𝑟, Find corresponding discrete value , Evaluate the objective function 𝑓𝑥𝑗
 Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants
 Update best path by: 𝜏𝑗
𝑛𝑒𝑤 ← 𝜏𝑗
𝑜𝑙𝑑 + ∆𝜏𝑗
𝑘
𝑘 and other paths by:𝜏𝑗
𝑛𝑒𝑤
← 1 − ρ 𝜏𝑗
𝑜𝑙𝑑
 If there is no convergence of the current solution & if 𝑡 > 𝑇 go to Loop
4. Print 𝑥𝑓𝑏𝑒𝑠𝑡 and 𝑓𝑏𝑒𝑠𝑡
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Pseudo code
1. Input
 Objective function (fitness function), upper bound (𝑢𝑏) and lower bound (𝑙𝑏), population size (𝑁), number of iteration 𝑇, scaling parameter 𝜁, evaporate rate ρ,
step size ℎ (or number of discrete value 𝑚)
2. Initialization
 Assume a set of permissible discrete values 𝑚 for each of the design variables
 Initialize all discrete values 𝑚 of design variables equal amounts of pheromone 𝜏
3. Loop:
 For 𝑡 = 1: 𝑇
 Find probability to select discrete values of design variables is 𝑝𝑗
𝑘
=
𝜏𝑗
𝜏𝑗
𝑚
𝑗=1
 Find the cumulative probability ranges associated with different discrete values based on its probabilities.(design roulette-wheel)
 For 𝑖 = 1: 𝑁 
 Generate a random numbers 𝑟, Find corresponding discrete value , Evaluate the objective function 𝑓𝑥𝑗
 Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants
 Update best path by: 𝜏𝑗
𝑛𝑒𝑤 ← 𝜏𝑗
𝑜𝑙𝑑 + ∆𝜏𝑗
𝑘
𝑘 and other paths by:𝜏𝑗
𝑛𝑒𝑤
← 1 − ρ 𝜏𝑗
𝑜𝑙𝑑
 If there is no convergence of the current solution & if 𝑡 > 𝑇 go to Loop
4. Print 𝑥𝑓𝑏𝑒𝑠𝑡 and 𝑓𝑏𝑒𝑠𝑡
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Pseudo code
1. Input
 Objective function (fitness function), upper bound (𝑢𝑏) and lower bound (𝑙𝑏), population size (𝑁), number of iteration 𝑇, scaling parameter 𝜁, evaporate rate ρ,
step size ℎ (or number of discrete value 𝑚)
2. Initialization
 Assume a set of permissible discrete values 𝑚 for each of the design variables
 Initialize all discrete values 𝑚 of design variables equal amounts of pheromone 𝜏
3. Loop:
 For 𝑡 = 1: 𝑇
 Find probability to select discrete values of design variables is 𝑝𝑗
𝑘
=
𝜏𝑗
𝜏𝑗
𝑚
𝑗=1
 Find the cumulative probability ranges associated with different discrete values based on its probabilities.(design roulette-wheel)
 For 𝑖 = 1: 𝑁 
 Generate a random numbers 𝑟, Find corresponding discrete value , Evaluate the objective function 𝑓𝑥𝑗
 Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants
 Update best path by: 𝜏𝑗
𝑛𝑒𝑤 ← 𝜏𝑗
𝑜𝑙𝑑 + ∆𝜏𝑗
𝑘
𝑘 and other paths by:𝜏𝑗
𝑛𝑒𝑤
← 1 − ρ 𝜏𝑗
𝑜𝑙𝑑
 If there is no convergence of the current solution & if 𝑡 > 𝑇 go to Loop
4. Print 𝑥𝑓𝑏𝑒𝑠𝑡 and 𝑓𝑏𝑒𝑠𝑡
Optimization Techniques
Swarm Optimization Techniques
 Ant Colony Optimization
• Pseudo code
1. Input
 Objective function (fitness function), upper bound (𝑢𝑏) and lower bound (𝑙𝑏), population size (𝑁), number of iteration 𝑇, scaling parameter 𝜁, evaporate rate ρ,
step size ℎ (or number of discrete value 𝑚)
2. Initialization
 Assume a set of permissible discrete values 𝑚 for each of the design variables
 Initialize all discrete values 𝑚 of design variables equal amounts of pheromone 𝜏
3. Loop:
 For 𝑡 = 1: 𝑇
 Find probability to select discrete values of design variables is 𝑝𝑗
𝑘
=
𝜏𝑗
𝜏𝑗
𝑚
𝑗=1
 Find the cumulative probability ranges associated with different discrete values based on its probabilities.(design roulette-wheel)
 For 𝑖 = 1: 𝑁 
 Generate a random numbers 𝑟, Find corresponding discrete value , Evaluate the objective function 𝑓𝑥𝑗
 Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants
 Update best path by: 𝜏𝑗
𝑛𝑒𝑤 ← 𝜏𝑗
𝑜𝑙𝑑 + ∆𝜏𝑗
𝑘
𝑘 and other paths by:𝜏𝑗
𝑛𝑒𝑤
← 1 − ρ 𝜏𝑗
𝑜𝑙𝑑
 If there is no convergence of the current solution & if 𝑡 > 𝑇 go to Loop
4. Print 𝑥𝑓𝑏𝑒𝑠𝑡 and 𝑓𝑏𝑒𝑠𝑡

Swarm Optimization Techniques_ACO.pdf

  • 1.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • A search technique used in computing to find near optimal solutions to discrete optimization problems. • ACO is a swarm intelligence inspired from the way that ants indirectly communicate directions to each other. • The most interesting aspect of the collaborative behaviour of ant species is their ability to find the shortest paths between the ants' nest and the food sources. • Finding the shortest path between their nest and a food source by chase pheromone trails exposed by other ants. The more intense the trail, the higher probability that an ant will follow it and thus enrich the trail with its own pheromone. The pheromone trails of longest paths evaporate.
  • 2.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • A search technique used in computing to find near optimal solutions to discrete optimization problems. • ACO is a swarm intelligence inspired from the way that ants indirectly communicate directions to each other. • The most interesting aspect of the collaborative behaviour of ant species is their ability to find the shortest paths between the ants' nest and the food sources. • Finding the shortest path between their nest and a food source by chase pheromone trails exposed by other ants. The more intense the trail, the higher probability that an ant will follow it and thus enrich the trail with its own pheromone. The pheromone trails of longest paths evaporate.
  • 3.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • A search technique used in computing to find near optimal solutions to discrete optimization problems. • ACO is a swarm intelligence inspired from the way that ants indirectly communicate directions to each other. • The most interesting aspect of the collaborative behaviour of ant species is their ability to find the shortest paths between the ants' nest and the food sources. • Finding the shortest path between their nest and a food source by chase pheromone trails exposed by other ants. The more intense the trail, the higher probability that an ant will follow it and thus enrich the trail with its own pheromone. The pheromone trails of longest paths evaporate.
  • 4.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • A search technique used in computing to find near optimal solutions to discrete optimization problems. • ACO is a swarm intelligence inspired from the way that ants indirectly communicate directions to each other. • The most interesting aspect of the collaborative behaviour of ant species is their ability to find the shortest paths between the ants' nest and the food sources. • Finding the shortest path between their nest and a food source by chase pheromone trails exposed by other ants. The more intense the trail, the higher probability that an ant will follow it and thus enrich the trail with its own pheromone. The pheromone trails of longest paths evaporate.
  • 5.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • A search technique used in computing to find near optimal solutions to discrete optimization problems. • ACO is a swarm intelligence inspired from the way that ants indirectly communicate directions to each other. • The most interesting aspect of the collaborative behaviour of ant species is their ability to find the shortest paths between the ants' nest and the food sources. • Finding the shortest path between their nest and a food source by chase pheromone trails exposed by other ants. The more intense the trail, the higher probability that an ant will follow it and thus enrich the trail with its own pheromone. The pheromone trails of longest paths evaporate.
  • 6.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Key Terms  Ants 𝑘: Any possible solution.  Population 𝑁- Group of all ants.  Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.  Search Space is divided by step size ℎ  Pheromone trail 𝜏  Scaling parameter 𝜁  Evaporate rate ρ
  • 7.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Key Terms  Ants 𝑘: Any possible solution.  Population 𝑁- Group of all ants.  Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.  Search Space is divided by step size ℎ  Pheromone trail 𝜏  Scaling parameter 𝜁  Evaporate rate ρ
  • 8.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Key Terms  Ants 𝑘: Any possible solution.  Population 𝑁- Group of all ants.  Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.  Search Space is divided by step size ℎ  Pheromone trail 𝜏  Scaling parameter 𝜁  Evaporate rate ρ
  • 9.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Key Terms  Ants 𝑘: Any possible solution.  Population 𝑁- Group of all ants.  Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.  Search Space is divided by step size ℎ  Pheromone trail 𝜏  Scaling parameter 𝜁  Evaporate rate ρ
  • 10.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Key Terms  Ants 𝑘: Any possible solution.  Population 𝑁- Group of all ants.  Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.  Search Space is divided by step size ℎ  Pheromone trail 𝜏  Scaling parameter 𝜁  Evaporate rate ρ
  • 11.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Key Terms  Ants 𝑘: Any possible solution.  Population 𝑁- Group of all ants.  Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.  Search Space is divided by step size ℎ  Pheromone trail 𝜏  Scaling parameter 𝜁  Evaporate rate ρ
  • 12.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Key Terms  Ants 𝑘: Any possible solution.  Population 𝑁- Group of all ants.  Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.  Search Space is divided by step size ℎ  Pheromone trail 𝜏  Scaling parameter 𝜁  Evaporate rate ρ
  • 13.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Key Terms  Ants 𝑘: Any possible solution.  Population 𝑁- Group of all ants.  Search Space [𝑙𝑏, 𝑢𝑏]- All possible solutions to the problem.  Search Space is divided by step size ℎ  Pheromone trail 𝜏  Scaling parameter 𝜁  Evaporate rate ρ
  • 14.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Initialization  Assume a suitable number of ants in the colony (population 𝑁)  Assume a set of permissible discrete values 𝑚 for each of the design variables (step size ℎ).  Initialize all discrete values of design variables equal amounts of pheromone 𝜏.
  • 15.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Initialization  Assume a suitable number of ants in the colony (population 𝑁)  Assume a set of permissible discrete values 𝑚 for each of the design variables (step size ℎ).  Initialize all discrete values of design variables equal amounts of pheromone 𝜏.
  • 16.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Initialization  Assume a suitable number of ants in the colony (population 𝑁)  Assume a set of permissible discrete values 𝑚 for each of the design variables (step size ℎ).  Initialize all discrete values of design variables equal amounts of pheromone 𝜏.
  • 17.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Initialization  Assume a suitable number of ants in the colony (population 𝑁)  Assume a set of permissible discrete values 𝑚 for each of the design variables (step size ℎ).  Initialize all discrete values of design variables equal amounts of pheromone 𝜏.
  • 18.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Build tours  From the home node, ants start travelling through the various paths and end at the destination node in each iteration (discrete values of design variables).  The probability to select discrete values of design variables is pj k = τj τj m j=1  Find the cumulative probability ranges associated with different discrete values based on its probabilities.  The specific discrete values chosen by ant k will be determined using the roulette-wheel selection. Roulette-wheel
  • 19.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Build tours  From the home node, ants start travelling through the various paths and end at the destination node in each iteration (discrete values of design variables).  The probability to select discrete values of design variables is pj k = τj τj m j=1  Find the cumulative probability ranges associated with different discrete values based on its probabilities.  The specific discrete values chosen by ant k will be determined using the roulette-wheel selection. Roulette-wheel
  • 20.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Build tours  From the home node, ants start travelling through the various paths and end at the destination node in each iteration (discrete values of design variables).  The probability to select discrete values of design variables is pj k = τj τj m j=1  Find the cumulative probability ranges associated with different discrete values based on its probabilities.  The specific discrete values chosen by ant k will be determined using the roulette-wheel selection. Roulette-wheel
  • 21.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Build tours  From the home node, ants start travelling through the various paths and end at the destination node in each iteration (discrete values of design variables).  The probability to select discrete values of design variables is pj k = τj τj m j=1  Find the cumulative probability ranges associated with different discrete values based on its probabilities.  The specific discrete values chosen by ant k will be determined using the roulette-wheel selection. Roulette-wheel
  • 22.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Build tours  From the home node, ants start travelling through the various paths and end at the destination node in each iteration (discrete values of design variables).  The probability to select discrete values of design variables is pj k = τj τj m j=1  Find the cumulative probability ranges associated with different discrete values based on its probabilities.  The specific discrete values chosen by ant k will be determined using the roulette-wheel selection. Roulette-wheel
  • 23.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 2 7 4 11 9 6 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel
  • 24.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel design variables
  • 25.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel 2 design variables
  • 26.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel 2 7 design variables
  • 27.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel 2 7 4 11 9 6 design variables
  • 28.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel 2 7 4 11 9 6 design variables cumulative probability for discrete values of design variables
  • 29.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel 2 7 4 11 9 6 design variables cumulative probability for discrete values of design variables 0.1
  • 30.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel 2 7 4 11 9 6 design variables cumulative probability for discrete values of design variables 0.1 0.3
  • 31.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel 2 7 4 11 9 6 design variables cumulative probability for discrete values of design variables 0.1 0.3 0.1 0.1 0.3 0.1
  • 32.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel 2 7 4 11 9 6 design variables cumulative probability for discrete values of design variables 0.1 0.3 0.1 0.1 0.3 0.1
  • 33.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel 2 7 4 11 9 6 design variables cumulative probability for discrete values of design variables 0.1 0.3 0.1 0.1 0.3 0.1
  • 34.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel 2 7 4 11 9 6 design variables cumulative probability for discrete values of design variables 0.1 0.3 0.1 0.1 0.3 0.1
  • 35.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel 2 7 4 11 9 6 design variables cumulative probability for discrete values of design variables 0.1 0.3 0.1 0.1 0.3 0.1 0.48
  • 36.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Select Path (solution)  Generate N random numbers r in the range (0, 1), one for each ant.  Determine the discrete value by ant k for variable as the one for which the cumulative probability range includes the random numbers r. 𝐴𝑛𝑡1 → 𝑟 = 0.48 → 𝑥 = 4 Roulette-wheel 2 7 𝟒 11 9 6 design variables cumulative probability for discrete values of design variables 0.1 0.3 0.1 0.1 0.3 0.1 0.48
  • 37.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Deposit and update trail  Evaluate the objective function values of each ant  Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants  Update the pheromone  Best ants: reinforcement the pheromone of the best path by: 𝜏𝑗 𝑛𝑒𝑤 ← 𝜏𝑗 𝑜𝑙𝑑 + ∆𝜏𝑗 𝑘 𝑘  Other ants: evaporates the pheromone of other paths by: 𝜏𝑗 𝑛𝑒𝑤 ← 1 − ρ 𝜏𝑗 𝑜𝑙𝑑 ∆𝜏𝑗 𝑘 = 𝜁𝑓𝑏𝑒𝑠𝑡 𝑓𝑤𝑜𝑟𝑠𝑡 𝜁 → scaling parameter 𝑓𝑏𝑒𝑠𝑡 → best objective function 𝑓𝑤𝑜𝑟𝑠𝑡 → worst best objective function ρ → evaporate rate
  • 38.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Deposit and update trail  Evaluate the objective function values of each ant  Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants  Update the pheromone  Best ants: reinforcement the pheromone of the best path by: 𝜏𝑗 𝑛𝑒𝑤 ← 𝜏𝑗 𝑜𝑙𝑑 + ∆𝜏𝑗 𝑘 𝑘  Other ants: evaporates the pheromone of other paths by: 𝜏𝑗 𝑛𝑒𝑤 ← 1 − ρ 𝜏𝑗 𝑜𝑙𝑑 ∆𝜏𝑗 𝑘 = 𝜁𝑓𝑏𝑒𝑠𝑡 𝑓𝑤𝑜𝑟𝑠𝑡 𝜁 → scaling parameter 𝑓𝑏𝑒𝑠𝑡 → best objective function 𝑓𝑤𝑜𝑟𝑠𝑡 → worst best objective function ρ → evaporate rate
  • 39.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Deposit and update trail  Evaluate the objective function values of each ant  Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants  Update the pheromone  Best ants: reinforcement the pheromone of the best path by: 𝜏𝑗 𝑛𝑒𝑤 ← 𝜏𝑗 𝑜𝑙𝑑 + ∆𝜏𝑗 𝑘 𝑘  Other ants: evaporates the pheromone of other paths by: 𝜏𝑗 𝑛𝑒𝑤 ← 1 − ρ 𝜏𝑗 𝑜𝑙𝑑 ∆𝜏𝑗 𝑘 = 𝜁𝑓𝑏𝑒𝑠𝑡 𝑓𝑤𝑜𝑟𝑠𝑡 𝜁 → scaling parameter 𝑓𝑏𝑒𝑠𝑡 → best objective function 𝑓𝑤𝑜𝑟𝑠𝑡 → worst best objective function ρ → evaporate rate
  • 40.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Deposit and update trail  Evaluate the objective function values of each ant  Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants  Update the pheromone  Best ants: reinforcement the pheromone of the best path by: 𝜏𝑗 𝑛𝑒𝑤 ← 𝜏𝑗 𝑜𝑙𝑑 + ∆𝜏𝑗 𝑘 𝑘  Other ants: evaporates the pheromone of other paths by: 𝜏𝑗 𝑛𝑒𝑤 ← 1 − ρ 𝜏𝑗 𝑜𝑙𝑑 ∆𝜏𝑗 𝑘 = 𝜁𝑓𝑏𝑒𝑠𝑡 𝑓𝑤𝑜𝑟𝑠𝑡 𝜁 → scaling parameter 𝑓𝑏𝑒𝑠𝑡 → best objective function 𝑓𝑤𝑜𝑟𝑠𝑡 → worst best objective function ρ → evaporate rate
  • 41.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Deposit and update trail  Evaluate the objective function values of each ant  Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants  Update the pheromone  Best ants: reinforcement the pheromone of the best path by: 𝜏𝑗 𝑛𝑒𝑤 ← 𝜏𝑗 𝑜𝑙𝑑 + ∆𝜏𝑗 𝑘 𝑘  Other ants: evaporates the pheromone of other paths by: 𝜏𝑗 𝑛𝑒𝑤 ← 1 − ρ 𝜏𝑗 𝑜𝑙𝑑 ∆𝜏𝑗 𝑘 = 𝜁𝑓𝑏𝑒𝑠𝑡 𝑓𝑤𝑜𝑟𝑠𝑡 𝜁 → scaling parameter 𝑓𝑏𝑒𝑠𝑡 → best objective function 𝑓𝑤𝑜𝑟𝑠𝑡 → worst best objective function ρ → evaporate rate
  • 42.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Deposit and update trail  Evaluate the objective function values of each ant  Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants  Update the pheromone  Best ants: reinforcement the pheromone of the best path by: 𝜏𝑗 𝑛𝑒𝑤 ← 𝜏𝑗 𝑜𝑙𝑑 + ∆𝜏𝑗 𝑘 𝑘  Other ants: evaporates the pheromone of other paths by: 𝜏𝑗 𝑛𝑒𝑤 ← 1 − ρ 𝜏𝑗 𝑜𝑙𝑑 ∆𝜏𝑗 𝑘 = 𝜁𝑓𝑏𝑒𝑠𝑡 𝑓𝑤𝑜𝑟𝑠𝑡 𝜁 → scaling parameter 𝑓𝑏𝑒𝑠𝑡 → best objective function 𝑓𝑤𝑜𝑟𝑠𝑡 → worst best objective function ρ → evaporate rate
  • 43.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Termination  The steps of ACO algorithm are iteratively repeated until the maximum number of iteration is reached or a termination criterion is met.  Convergence: is the case where the positions of all particles converge to the same set of values, the method is assumed to have converged.
  • 44.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Termination  The steps of ACO algorithm are iteratively repeated until the maximum number of iteration is reached or a termination criterion is met.  Convergence: is the case where the positions of all particles converge to the same set of values, the method is assumed to have converged.
  • 45.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Procedure  Termination  The steps of ACO algorithm are iteratively repeated until the maximum number of iteration is reached or a termination criterion is met.  Convergence: is the case where the of all ants converge to the same set of values, the method is assumed to have converged.
  • 46.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Parameters required from user:  Population size 𝑁  Set of permissible discrete values 𝑚 for each of the design variables  Step size ℎ  Initial pheromone trail 𝜏  Scaling parameter 𝜁  Evaporate rate ρ  Termination criteria (i.e. number of iteration 𝑇)
  • 47.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Pseudo code 1. Input  Objective function (fitness function), upper bound (𝑢𝑏) and lower bound (𝑙𝑏), population size (𝑁), number of iteration 𝑇, scaling parameter 𝜁, evaporate rate ρ, step size ℎ (or number of discrete value 𝑚) 2. Initialization  Assume a set of permissible discrete values 𝑚 for each of the design variables  Initialize all discrete values 𝑚 of design variables equal amounts of pheromone 𝜏 3. Loop:  For 𝑡 = 1: 𝑇  Find probability to select discrete values of design variables is 𝑝𝑗 𝑘 = 𝜏𝑗 𝜏𝑗 𝑚 𝑗=1  Find the cumulative probability ranges associated with different discrete values based on its probabilities.(design roulette-wheel)  For 𝑖 = 1: 𝑁   Generate a random numbers 𝑟, Find corresponding discrete value , Evaluate the objective function 𝑓𝑥𝑗  Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants  Update best path by: 𝜏𝑗 𝑛𝑒𝑤 ← 𝜏𝑗 𝑜𝑙𝑑 + ∆𝜏𝑗 𝑘 𝑘 and other paths by:𝜏𝑗 𝑛𝑒𝑤 ← 1 − ρ 𝜏𝑗 𝑜𝑙𝑑  If there is no convergence of the current solution & if 𝑡 > 𝑇 go to Loop 4. Print 𝑥𝑓𝑏𝑒𝑠𝑡 and 𝑓𝑏𝑒𝑠𝑡
  • 48.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Pseudo code 1. Input  Objective function (fitness function), upper bound (𝑢𝑏) and lower bound (𝑙𝑏), population size (𝑁), number of iteration 𝑇, scaling parameter 𝜁, evaporate rate ρ, step size ℎ (or number of discrete value 𝑚) 2. Initialization  Assume a set of permissible discrete values 𝑚 for each of the design variables  Initialize all discrete values 𝑚 of design variables equal amounts of pheromone 𝜏 3. Loop:  For 𝑡 = 1: 𝑇  Find probability to select discrete values of design variables is 𝑝𝑗 𝑘 = 𝜏𝑗 𝜏𝑗 𝑚 𝑗=1  Find the cumulative probability ranges associated with different discrete values based on its probabilities.(design roulette-wheel)  For 𝑖 = 1: 𝑁   Generate a random numbers 𝑟, Find corresponding discrete value , Evaluate the objective function 𝑓𝑥𝑗  Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants  Update best path by: 𝜏𝑗 𝑛𝑒𝑤 ← 𝜏𝑗 𝑜𝑙𝑑 + ∆𝜏𝑗 𝑘 𝑘 and other paths by:𝜏𝑗 𝑛𝑒𝑤 ← 1 − ρ 𝜏𝑗 𝑜𝑙𝑑  If there is no convergence of the current solution & if 𝑡 > 𝑇 go to Loop 4. Print 𝑥𝑓𝑏𝑒𝑠𝑡 and 𝑓𝑏𝑒𝑠𝑡
  • 49.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Pseudo code 1. Input  Objective function (fitness function), upper bound (𝑢𝑏) and lower bound (𝑙𝑏), population size (𝑁), number of iteration 𝑇, scaling parameter 𝜁, evaporate rate ρ, step size ℎ (or number of discrete value 𝑚) 2. Initialization  Assume a set of permissible discrete values 𝑚 for each of the design variables  Initialize all discrete values 𝑚 of design variables equal amounts of pheromone 𝜏 3. Loop:  For 𝑡 = 1: 𝑇  Find probability to select discrete values of design variables is 𝑝𝑗 𝑘 = 𝜏𝑗 𝜏𝑗 𝑚 𝑗=1  Find the cumulative probability ranges associated with different discrete values based on its probabilities.(design roulette-wheel)  For 𝑖 = 1: 𝑁   Generate a random numbers 𝑟, Find corresponding discrete value , Evaluate the objective function 𝑓𝑥𝑗  Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants  Update best path by: 𝜏𝑗 𝑛𝑒𝑤 ← 𝜏𝑗 𝑜𝑙𝑑 + ∆𝜏𝑗 𝑘 𝑘 and other paths by:𝜏𝑗 𝑛𝑒𝑤 ← 1 − ρ 𝜏𝑗 𝑜𝑙𝑑  If there is no convergence of the current solution & if 𝑡 > 𝑇 go to Loop 4. Print 𝑥𝑓𝑏𝑒𝑠𝑡 and 𝑓𝑏𝑒𝑠𝑡
  • 50.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Pseudo code 1. Input  Objective function (fitness function), upper bound (𝑢𝑏) and lower bound (𝑙𝑏), population size (𝑁), number of iteration 𝑇, scaling parameter 𝜁, evaporate rate ρ, step size ℎ (or number of discrete value 𝑚) 2. Initialization  Assume a set of permissible discrete values 𝑚 for each of the design variables  Initialize all discrete values 𝑚 of design variables equal amounts of pheromone 𝜏 3. Loop:  For 𝑡 = 1: 𝑇  Find probability to select discrete values of design variables is 𝑝𝑗 𝑘 = 𝜏𝑗 𝜏𝑗 𝑚 𝑗=1  Find the cumulative probability ranges associated with different discrete values based on its probabilities.(design roulette-wheel)  For 𝑖 = 1: 𝑁   Generate a random numbers 𝑟, Find corresponding discrete value , Evaluate the objective function 𝑓𝑥𝑗  Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants  Update best path by: 𝜏𝑗 𝑛𝑒𝑤 ← 𝜏𝑗 𝑜𝑙𝑑 + ∆𝜏𝑗 𝑘 𝑘 and other paths by:𝜏𝑗 𝑛𝑒𝑤 ← 1 − ρ 𝜏𝑗 𝑜𝑙𝑑  If there is no convergence of the current solution & if 𝑡 > 𝑇 go to Loop 4. Print 𝑥𝑓𝑏𝑒𝑠𝑡 and 𝑓𝑏𝑒𝑠𝑡
  • 51.
    Optimization Techniques Swarm OptimizationTechniques  Ant Colony Optimization • Pseudo code 1. Input  Objective function (fitness function), upper bound (𝑢𝑏) and lower bound (𝑙𝑏), population size (𝑁), number of iteration 𝑇, scaling parameter 𝜁, evaporate rate ρ, step size ℎ (or number of discrete value 𝑚) 2. Initialization  Assume a set of permissible discrete values 𝑚 for each of the design variables  Initialize all discrete values 𝑚 of design variables equal amounts of pheromone 𝜏 3. Loop:  For 𝑡 = 1: 𝑇  Find probability to select discrete values of design variables is 𝑝𝑗 𝑘 = 𝜏𝑗 𝜏𝑗 𝑚 𝑗=1  Find the cumulative probability ranges associated with different discrete values based on its probabilities.(design roulette-wheel)  For 𝑖 = 1: 𝑁   Generate a random numbers 𝑟, Find corresponding discrete value , Evaluate the objective function 𝑓𝑥𝑗  Determine the best 𝑓𝑏𝑒𝑠𝑡 and worst 𝑓𝑤𝑜𝑟𝑠𝑡 objective function of the discrete value among ants  Update best path by: 𝜏𝑗 𝑛𝑒𝑤 ← 𝜏𝑗 𝑜𝑙𝑑 + ∆𝜏𝑗 𝑘 𝑘 and other paths by:𝜏𝑗 𝑛𝑒𝑤 ← 1 − ρ 𝜏𝑗 𝑜𝑙𝑑  If there is no convergence of the current solution & if 𝑡 > 𝑇 go to Loop 4. Print 𝑥𝑓𝑏𝑒𝑠𝑡 and 𝑓𝑏𝑒𝑠𝑡