Swarm Intelligence Flashcards

(63 cards)

1
Q

One property of a swarm is each member’s inability to do a task by itself, but…

A

It can do the same task easily as a swarm

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

The ability to do a task that a member could not do otherwise is called…

A

An emergent property of the swarm

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

Each element of a swarm has…

A

A simple behaviour, and a set of rules for interacting with other members and the environment

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Ant Colony Optimisation is an evolutionary algorithm inspired by ants, used for the task of…

A

Finding the shortest path/optimisation of a task (both are technically true)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Stimergy is a phenomenon we observe in nature, such that…

A

Animals communicate via interaction with the environment

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Sematonic stimergy is where…

A

The action of an agent is directly related to problem solving

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Sign-based stimergy is where…

A

The action of an agent affects the environment, but is not directly aimed at problem solving

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Ants have highly sophisticated [sematonic/sign-based] stimergy, where they communicate via […].

A

Sign-based, pheromones

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

The stronger scented a pheromone trail is, the more likely an ant has been down this path before and [failed to return/returned safely], therefore it is [less/more] likely for an ant to take this path.

A

Returned safely, more

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

In ACO, ants are […] that move along […] between nodes in a graph.

A

Agents, edges

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

In ACO, an ant’s path represents a specific…

A

Candidate solution

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

In ACO, after an ant has generated a candidate solution…

A

Pheromone is laid on that path

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

In ACO, [more/less] pheromones are placed on a path based on the [ant/quality of the solution].

A

More, based on the quality of the solution

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

In ACO, the initial graph begins with [no/equal/random] pheromone on along each edge.

A

Random

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

In ACO, an ant does not return to a node it has already been to due to…

A

A tabulist describing the nodes it has already visited

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

In ACO, after each [ant/iteration], we evaporate the pheromone.

A

Iteration

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

In ACO, we define the probability that an ant will go to a city as…

A

The probability we are going to that city, divided by the probability we will go to the other cities

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

In ACO, pheromones are updated using the update rule, which measures each update as…

A

The original pheromone multiplied by a learning rate coefficient, plus some measure of the quality of the path

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

The Max-Min Ant System updates pheromone trails based on […] ants.

A

The best

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

The Elitist Rank Ant System updates pheromone trails based on […] ants.

A

The best of a set number of

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

Evolutionary algorithms like ACO can be applied to any problem as long as you have…

A

Relevant embedding

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

A construction graph represents…

A

The construction of the problem, where we represent each decision horizontally as a node linked to each possible remaining choice

For example, given a graph of size 10, we start at the beginning node, then make some decision of the 9 remaining nodes. This becomes a new node, which is connected to 8 nodes, and so on.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

Swarm intelligence is defined as…

A

Emergent intelligent behaviour not possible by a single individual, but becomes possible as a group

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
24
Q

Collective movement is when…

A

A swarm moves as a collective in some way

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
25
An example of collective, coordinated movement is... (pick one)
Flocking, swarming, formation travelling
26
Movement is not considered collective if the movement is...
Movement that is pre-determined before the animal sets off or fixed movement, as it is not considered intelligent movement
27
Flocking is a type of collective movement wherein the group exhibits...
Fast, directed movement with no leader that can react to predators and obstacles, while avoiding collision between flock members
28
Flocking exhibits a balance between [...] and [...].
Attraction and repulsion
29
Geese fly in a V-formation when migrating, in order to improve...
Aerodynamism
30
Slime cast off by fish as they swim reduces turbulence in water, improving...
The consumption of energy of the fish behind them
31
Flash expansion is a phenomenon observed in prey, wherein prey...
Dashes in all directions at once, in order to confuse the predator
32
Boids are a type of algorithm that aims to simulate...
Flocking behaviours
33
Each boid has a [...] that extends from the front of each boid.
Neighbourhood/vision cone
34
The separation rule of boids states that a boid must...
Steer to avoid crowding local flockmates
35
The alignment rule of boids states that a boid must...
Steer towards the average heading of the flockmates
36
The cohesion rule of boids states that a boid must...
Steer towards the average position of the local flockmates
37
If boids are started too close together, we may encounter...
Flash expansion
38
If boids are started too far apart, we may encounter...
Slow aggregation, or small subflocks
39
Particle Swarm Optimisation is an evolutionary algorithm inspired by flocking behaviour, used for the task of...
Finding the optimal set of continuous values for a task
40
In PSO, each candidate solution is encoded as a...
Particle
41
In PSO, the candidate space is a...
Swarm of particles
42
In PSO, particles are physics objects that are only affected by [...] and [...].
Velocity and acceleration
43
Particle Swarm Optimisation is a [competitive/collaborative] multi-agent system.
Collaborative
44
In PSO, we begin by...
Initialising a number of random particles along the search space
45
In PSO, for each timestep, we...
Update the velocity of each particle, then using that value to update position via addition
46
In PSO, we update velocity by...
Taking the current velocity plus the distance to a random weighting of the particle's best encountered position and a random weighting of the distance to the population's best encountered position
47
In PSO, besides max iterations and a lack of improvement, we terminate when...
The normalised swarm radius is close to zero, meaning the swarm has converged
48
PSO can be updated to work on discrete spaces y...
Updating positions using the velocity as a probability threshold instead of an additive value
49
An archive is a term used in multi-objective algorithms to describe...
The set of solutions described by a Pareto front approximation at a given timestep
50
If a new solution dominates one of the solutions in a Pareto front archive...
We remove that solution and replace it with our new one
51
If a new solution does not dominate one of the solutions in a Pareto front archive...
We mark it as dominated, then continue onto the next solution and evaluate again
52
If a new solution does not dominate any of the solutions in a Pareto front archive...
We do not add it to the archive
53
If a new solution does not dominate any of the solutions in a Pareto front archive, but is also not marked as dominated...
The archive is likely empty, hence we add it anyway
54
It is difficult to find leaders in a large Pareto front archive because...
It is computationally expensive to do so
55
Diversity preservation must be used in Pareto front archives to...
Preserve diversity in the set of solutions
56
Clustering is a form of diversity preservation used in a Pareto front archive by...
Separating sets of solutions into clusters of solutions, keeping the solution with the minimum average distance to the cluster center
57
Crowding distance is a form of diversity preservation used in a Pareto front archive by...
Scoring solutions according to their proximity to their neighbours, giving higher scores based on how far apart they are
58
Leader selection must be used in Pareto front archives to...
Attract the swarm towards the set of non-dominated solutions on the Pareto front
59
Random selection is used for leader selection in Pareto front archives by...
Selecting a random archive member, either truly randomly or using some weighted selection
60
Dominated solution count is used for leader selection in Pareto front archives by...
Ranking the solutions in the archive based on the number of solutions they dominate
61
Hypervolumes are used for leader selection in Pareto front archives by...
Picking the solution that has the largest hypervolume (the area it dominates between it and a given reference point)
62
Many-objective PSO can be ranked using a ranking method like average rank, which...
Ranks the archived solution once for each objective, and takes the average of these ranks
63
Many-objective PSO can be ranked using a ranking method like distance ranking, which...
Defines the rank of a solution based on the distance between each objective value