Table of Contents

Class ParticleSwarmOptimizationOptions<T, TInput, TOutput>

Namespace
AiDotNet.Models.Options
Assembly
AiDotNet.dll

Configuration options for Particle Swarm Optimization (PSO), a population-based stochastic optimization technique inspired by social behavior of bird flocking or fish schooling.

public class ParticleSwarmOptimizationOptions<T, TInput, TOutput> : OptimizationAlgorithmOptions<T, TInput, TOutput>

Type Parameters

T
TInput
TOutput
Inheritance
OptimizationAlgorithmOptions<T, TInput, TOutput>
ParticleSwarmOptimizationOptions<T, TInput, TOutput>
Inherited Members

Remarks

Particle Swarm Optimization is a computational method that optimizes a problem by iteratively improving candidate solutions (particles) with regard to a given measure of quality. The algorithm maintains a population of particles, where each particle represents a potential solution to the optimization problem. Particles move through the solution space guided by their own best known position and the swarm's best known position. This social interaction leads to emergent intelligence that efficiently explores complex solution spaces. PSO is particularly effective for continuous optimization problems and has advantages in terms of simplicity, flexibility, and minimal parameter tuning requirements.

For Beginners: Particle Swarm Optimization is a way to find the best solution by mimicking how birds flock or fish swim in groups.

Imagine you and your friends are searching for the highest point in a hilly landscape while blindfolded:

  • Each person (particle) can only feel the height at their current position
  • Everyone remembers the highest point they personally have found so far
  • The group also shares information about the highest point anyone has found

When deciding where to step next:

  • You consider continuing in your current direction (inertia)
  • You're pulled toward the best spot you've personally found (cognitive component)
  • You're also pulled toward the best spot anyone in the group has found (social component)
  • You combine these influences to decide your next move

This creates a smart search pattern because:

  • People explore different areas (maintaining diversity in the search)
  • The group gradually converges on promising regions
  • The method balances exploration (finding new areas) with exploitation (refining good solutions)

This class provides extensive options to fine-tune how the particles move and interact, allowing you to customize the algorithm for different types of optimization problems.

Properties

CognitiveParameter

Gets or sets the cognitive parameter that controls the influence of the particle's personal best position.

public double CognitiveParameter { get; set; }

Property Value

double

The cognitive parameter, defaulting to 1.49445.

Remarks

The cognitive parameter, also known as c1 or the personal acceleration coefficient, determines how much the particle is influenced by its own past best position. This coefficient controls the "nostalgia" of the particle, encouraging it to return to regions where it personally found good solutions. Higher values increase the attraction toward the particle's personal best position. The default value of 1.49445, when used with the corresponding social parameter and inertia weight, comes from Clerc's constriction coefficient approach, which provides theoretically backed convergence properties.

For Beginners: This setting controls how strongly a particle is attracted to the best position it has personally discovered.

The default value of 1.49445 means:

  • The particle is strongly influenced by its memory of the best spot it has found
  • This creates a balance between exploring new areas and revisiting proven good spots

Think of it like your own experience when searching:

  • High cognitive value: You strongly trust your own experience and return to places you found valuable
  • Low cognitive value: You're less attached to your past discoveries and more willing to try new directions

You might want a higher value if:

  • You want particles to more thoroughly explore areas they've found promising
  • You want to emphasize independent exploration over group consensus
  • The problem has many local optima where personal experience is valuable

You might want a lower value if:

  • You want particles to be less fixated on their own discoveries
  • You prefer faster convergence toward the group's best find
  • You have many particles and want more coordinated movement

The specific default value (1.49445) works in harmony with the default inertia and social parameter to provide mathematically proven convergence properties.

CognitiveWeightAdaptationRate

Gets or sets the rate at which the cognitive weight adapts when using adaptive weights.

public double CognitiveWeightAdaptationRate { get; set; }

Property Value

double

The cognitive weight adaptation rate, defaulting to 1.0.

Remarks

This parameter controls how rapidly the cognitive parameter changes when UseAdaptiveWeights is enabled. Higher values lead to faster adaptation, while lower values result in more gradual changes. The exact effect depends on the specific adaptive weight scheme implemented, but generally this acts as a multiplier on the amount of change applied to the cognitive parameter in each iteration. This parameter is only relevant when UseAdaptiveWeights is set to true.

For Beginners: This setting controls how quickly the cognitive weight (attraction to personal best) changes over time when adaptive weights are enabled.

The default value of 1.0 means:

  • The cognitive weight changes at a standard rate
  • Neither accelerated nor slowed down

Think of it like setting the speed for changing from individual exploration to group consensus:

  • Higher values (like 1.5) make this transition happen faster
  • Lower values (like 0.5) make this transition happen more gradually

You might want a higher value if:

  • You want faster transition from exploration to exploitation
  • You have limited iterations available for the search
  • You want more responsive adaptation to the optimization progress

You might want a lower value if:

  • You want more gradual, stable changes in search behavior
  • You have many iterations available
  • You want the exploration phase to last longer

This setting has no effect unless UseAdaptiveWeights is set to true.

InertiaDecayRate

Gets or sets the rate at which inertia weight decays when using adaptive inertia.

public double InertiaDecayRate { get; set; }

Property Value

double

The inertia decay rate, defaulting to 0.99.

Remarks

This parameter controls how rapidly the inertia weight changes when UseAdaptiveInertia is enabled. In typical implementations, this represents a multiplicative factor applied to the inertia weight after each iteration. Values closer to 1.0 result in slower decay and more gradual transition from exploration to exploitation. Values further from 1.0 (smaller) cause faster decay and more rapid transition. This parameter is only relevant when UseAdaptiveInertia is set to true.

For Beginners: This setting controls how quickly the inertia weight changes over time when adaptive inertia is enabled.

The default value of 0.99 means:

  • After each iteration, the inertia weight is multiplied by 0.99
  • This creates a very gradual reduction in momentum over time

Think of it like gradually slowing down as the search progresses:

  • A value near 1.0 (like 0.99) creates a very slow, gradual transition
  • A smaller value (like 0.95) creates a much faster transition from exploration to exploitation

You might want a value closer to 1.0 (like 0.995) if:

  • You want a very gradual transition from exploration to exploitation
  • You have many iterations available for the search
  • You want to maintain exploration capability for longer

You might want a smaller value (like 0.95) if:

  • You want a quicker transition to focused refinement
  • You have limited iterations available
  • You want the algorithm to settle faster

This setting has no effect unless UseAdaptiveInertia is set to true.

InertiaWeight

Gets or sets the inertia weight that controls the influence of the particle's previous velocity.

public double InertiaWeight { get; set; }

Property Value

double

The inertia weight, defaulting to 0.729.

Remarks

The inertia weight controls how much of the particle's previous velocity is retained when updating its position. Higher values (near 1.0) favor exploration by maintaining momentum, allowing particles to overshoot optima and continue exploring new regions. Lower values prioritize exploitation of known good regions. The default value of 0.729 is commonly used in conjunction with acceleration coefficients of 1.49445 (cognitive and social parameters) as it provides guaranteed convergence in simplified systems. This parameterization is known as Clerc's constriction coefficient approach.

For Beginners: This setting determines how much a particle tends to continue moving in its current direction.

The default value of 0.729 means:

  • About 73% of the particle's previous movement direction is maintained
  • This creates a balance between exploring new areas and refining current good spots

Think of it like momentum when walking:

  • High inertia (like 0.9): You tend to keep going in the same direction, even if you see something interesting nearby
  • Low inertia (like 0.4): You can quickly change direction toward interesting areas

You might want higher inertia if:

  • Your problem has a complex landscape where thorough exploration is needed
  • The algorithm seems to get stuck in poor solutions
  • You want to prioritize exploration over exploitation

You might want lower inertia if:

  • You want faster convergence to good solutions
  • The algorithm seems to bounce around too much without settling
  • You want to prioritize refinement of promising solutions

The default value (0.729) is mathematically derived to provide good convergence properties when used with the default cognitive and social parameters (1.49445).

InitialCognitiveWeight

Gets or sets the initial cognitive weight when using adaptive weights.

public double InitialCognitiveWeight { get; set; }

Property Value

double

The initial cognitive weight, defaulting to 1.5.

Remarks

This parameter defines the starting value for the cognitive parameter when UseAdaptiveWeights is enabled. The cognitive parameter controls how much a particle is influenced by its own past best position. In adaptive schemes, this weight typically starts higher to promote exploration based on personal experience, and then decreases to favor social learning in later stages. This parameter is only relevant when UseAdaptiveWeights is set to true.

For Beginners: This setting determines the starting value for how strongly particles are attracted to their own personal best positions when adaptive weights are enabled.

The default value of 1.5 means:

  • Particles start with a relatively strong attraction to their personal best findings
  • This encourages initial independent exploration based on individual experience

Think of it like setting initial confidence in one's own discoveries:

  • Higher values make searchers more likely to revisit and explore around their own best finds
  • Lower values make searchers less attached to their previous discoveries

You might want a higher value (like 2.0) if:

  • You want more diverse, independent exploration early in the search
  • You suspect multiple promising regions that should be investigated thoroughly
  • You want to delay convergence to maintain diversity

You might want a lower value (like 1.0) if:

  • You want more coordinated exploration from the beginning
  • You have many particles and want to avoid too much redundant searching
  • You prefer faster initial convergence toward promising areas

This setting has no effect unless UseAdaptiveWeights is set to true.

InitialInertia

Gets or sets the initial inertia weight when using adaptive inertia.

public double InitialInertia { get; set; }

Property Value

double

The initial inertia weight, defaulting to 0.7.

Remarks

This parameter defines the starting inertia weight when UseAdaptiveInertia is enabled. The inertia weight will gradually change from this initial value during the optimization process, typically decreasing to favor exploitation in later iterations. This parameter is only relevant when UseAdaptiveInertia is set to true. A moderate initial value allows particles to begin with a balance between exploration and exploitation, with the balance shifting as the optimization progresses.

For Beginners: This setting determines the starting value for the inertia weight when adaptive inertia is enabled.

The default value of 0.7 means:

  • Particles start with 70% of their previous velocity maintained
  • This provides a moderate level of momentum for initial exploration

Think of it like setting the initial momentum for your search team:

  • Higher values (like 0.9) create more initial momentum, favoring broad exploration
  • Lower values (like 0.5) create less initial momentum, favoring more directed movement

You might want a higher value if:

  • Your problem requires thorough exploration of the search space
  • You want to avoid premature convergence to local optima
  • Your landscape has many hills and valleys to navigate

You might want a lower value if:

  • You have a good initial estimate of where the solution lies
  • You want faster initial convergence toward promising regions
  • Your problem is relatively well-behaved with fewer local optima

This setting has no effect unless UseAdaptiveInertia is set to true.

InitialSocialWeight

Gets or sets the initial social weight when using adaptive weights.

public double InitialSocialWeight { get; set; }

Property Value

double

The initial social weight, defaulting to 1.5.

Remarks

This parameter defines the starting value for the social parameter when UseAdaptiveWeights is enabled. The social parameter controls how much a particle is influenced by the best position found by any particle in the swarm. In adaptive schemes, this weight typically starts lower to reduce premature convergence, and then increases to promote convergence in later stages. This parameter is only relevant when UseAdaptiveWeights is set to true.

For Beginners: This setting determines the starting value for how strongly particles are attracted to the swarm's global best position when adaptive weights are enabled.

The default value of 1.5 means:

  • Particles start with a moderate attraction to the group's best find
  • This balances independent exploration with some group coordination

Think of it like setting initial trust in the group's collective knowledge:

  • Higher values make searchers more likely to focus on the best area found by anyone
  • Lower values make searchers more independent from the group consensus

You might want a higher value (like 2.0) if:

  • You want faster initial convergence toward promising areas
  • You have confidence in the early discoveries being near the true optimum
  • You want more coordinated group behavior from the start

You might want a lower value (like 1.0) if:

  • You want to prevent premature convergence to potentially suboptimal solutions
  • You want to ensure thorough exploration before the swarm begins to converge
  • You're dealing with a problem with many local optima that could mislead the search

This setting has no effect unless UseAdaptiveWeights is set to true.

MaxCognitiveWeight

Gets or sets the maximum cognitive weight when using adaptive weights.

public double MaxCognitiveWeight { get; set; }

Property Value

double

The maximum cognitive weight, defaulting to 2.5.

Remarks

This parameter defines the upper bound for the cognitive parameter when UseAdaptiveWeights is enabled. The adaptive weight scheme will not increase the cognitive parameter above this value, preventing particles from becoming excessively attracted to their personal best positions. This helps maintain stable search behavior by limiting the cognitive component's influence. This parameter is only relevant when UseAdaptiveWeights is set to true and is particularly important in schemes that might dynamically adjust weights based on swarm performance.

For Beginners: This setting sets an upper limit for how high the cognitive weight can increase when adaptive weights are enabled.

The default value of 2.5 means:

  • The attraction to personal best positions will never exceed this level
  • This prevents particles from becoming overly fixated on their own discoveries

Think of it like setting a maximum level of confidence in your own findings:

  • Prevents searchers from being so attached to their own discoveries that they ignore others
  • Keeps the balance between personal experience and group knowledge

You might want a higher value (like 3.0) if:

  • You want to strongly emphasize individual exploration
  • You want particles to thoroughly investigate their personal best regions
  • Your problem benefits from very diverse, independent search behavior

You might want a lower value (like 2.0) if:

  • You want more moderate, balanced attraction forces
  • You're concerned about numerical stability with high acceleration
  • You prefer more coordination between particles

This setting has no effect unless UseAdaptiveWeights is set to true.

MaxInertia

Gets or sets the maximum inertia weight when using adaptive inertia.

public double MaxInertia { get; set; }

Property Value

double

The maximum inertia weight, defaulting to 0.9.

Remarks

This parameter defines the upper bound for the inertia weight when UseAdaptiveInertia is enabled. The adaptive inertia scheme will not increase the inertia above this value, preventing particles from gaining excessive momentum that might cause them to overshoot promising regions too drastically. This helps maintain stable exploration behavior. This parameter is only relevant when UseAdaptiveInertia is set to true and is particularly important in schemes that might dynamically increase inertia based on swarm performance.

For Beginners: This setting sets an upper limit for how high the inertia weight can increase when adaptive inertia is enabled.

The default value of 0.9 means:

  • The inertia weight will never exceed 90% of previous velocity
  • This prevents particles from gaining too much momentum

Think of it like setting a speed limit for your search:

  • Prevents searchers from moving so quickly they miss important details
  • Keeps the search behavior stable and controllable

You might want a higher value (like 0.95) if:

  • Your search space is very large and requires extensive exploration
  • You want particles to more quickly traverse large regions
  • Your problem has wide, flat areas that need to be crossed efficiently

You might want a lower value (like 0.8) if:

  • You want more controlled, cautious exploration
  • Your problem has closely spaced features that require careful examination
  • You're concerned about numerical stability with high momentum

This setting has no effect unless UseAdaptiveInertia is set to true.

MaxSocialWeight

Gets or sets the maximum social weight when using adaptive weights.

public double MaxSocialWeight { get; set; }

Property Value

double

The maximum social weight, defaulting to 2.5.

Remarks

This parameter defines the upper bound for the social parameter when UseAdaptiveWeights is enabled. The adaptive weight scheme will not increase the social parameter above this value, preventing particles from becoming excessively attracted to the global best position. This helps avoid premature convergence and maintains diversity in the swarm. This parameter is only relevant when UseAdaptiveWeights is set to true and is particularly important in schemes that increase social influence in later stages.

For Beginners: This setting sets an upper limit for how high the social weight can increase when adaptive weights are enabled.

The default value of 2.5 means:

  • The attraction to the swarm's best position will never exceed this level
  • This prevents excessive focus on a single promising solution

Think of it like setting a maximum level of group influence:

  • Prevents searchers from blindly rushing to the current best location
  • Maintains some individuality even when converging toward good solutions

You might want a higher value (like 3.0) if:

  • You want stronger convergence in later stages
  • You want to more aggressively refine the best found solution
  • Your problem benefits from focused exploitation once good regions are identified

You might want a lower value (like 2.0) if:

  • You want more balanced attraction forces
  • You're concerned about premature convergence
  • You prefer more gradual convergence behavior

This setting has no effect unless UseAdaptiveWeights is set to true.

MinCognitiveWeight

Gets or sets the minimum cognitive weight when using adaptive weights.

public double MinCognitiveWeight { get; set; }

Property Value

double

The minimum cognitive weight, defaulting to 0.5.

Remarks

This parameter defines the lower bound for the cognitive parameter when UseAdaptiveWeights is enabled. The adaptive weight scheme will not reduce the cognitive parameter below this value, ensuring that particles maintain some level of attraction to their personal best positions even in later stages. This helps prevent complete abandonment of personally discovered promising regions. This parameter is only relevant when UseAdaptiveWeights is set to true.

For Beginners: This setting sets a lower limit for how much the cognitive weight can decrease when adaptive weights are enabled.

The default value of 0.5 means:

  • The attraction to personal best positions will never fall below this level
  • This ensures particles always maintain some memory of their own discoveries

Think of it like maintaining a minimum level of confidence in your own findings:

  • Even when the group has found good solutions, you still value your own experience
  • This prevents completely abandoning your own discoveries

You might want a lower value (like 0.2) if:

  • You want stronger convergence behavior in later stages
  • You believe the global best is likely to be near the true optimum
  • You want particles to eventually focus almost exclusively on refinement

You might want a higher value (like 0.8) if:

  • You want to maintain diversity throughout the search
  • You're concerned about premature convergence to suboptimal solutions
  • Your problem has many promising regions that should be continuously explored

This setting has no effect unless UseAdaptiveWeights is set to true.

MinInertia

Gets or sets the minimum inertia weight when using adaptive inertia.

public double MinInertia { get; set; }

Property Value

double

The minimum inertia weight, defaulting to 0.1.

Remarks

This parameter defines the lower bound for the inertia weight when UseAdaptiveInertia is enabled. The adaptive inertia scheme will not reduce the inertia below this value, ensuring that particles maintain at least some momentum even in later stages of optimization. This prevents the swarm from becoming completely static and allows for continued refinement of solutions. This parameter is only relevant when UseAdaptiveInertia is set to true.

For Beginners: This setting sets a lower limit for how much the inertia weight can decrease when adaptive inertia is enabled.

The default value of 0.1 means:

  • The inertia weight will never drop below 10% of previous velocity
  • This ensures particles always maintain some degree of momentum

Think of it like setting a minimum speed for your search:

  • Even in late stages, searchers won't come to a complete stop
  • This prevents getting permanently stuck in a single location

You might want a lower value (like 0.05) if:

  • You want very fine-grained refinement of solutions in later stages
  • Your problem benefits from very small, careful steps near the optimum
  • You have many iterations to allow for gradual convergence

You might want a higher value (like 0.3) if:

  • You want to maintain more exploration capability throughout
  • You're concerned about getting stuck in local optima
  • Your problem landscape has many closely spaced local optima

This setting has no effect unless UseAdaptiveInertia is set to true.

MinSocialWeight

Gets or sets the minimum social weight when using adaptive weights.

public double MinSocialWeight { get; set; }

Property Value

double

The minimum social weight, defaulting to 0.5.

Remarks

This parameter defines the lower bound for the social parameter when UseAdaptiveWeights is enabled. The adaptive weight scheme will not reduce the social parameter below this value, ensuring that particles maintain some level of attraction to the global best position even in early stages of optimization. This helps prevent complete isolation of particles and enables some level of coordination within the swarm. This parameter is only relevant when UseAdaptiveWeights is set to true.

For Beginners: This setting sets a lower limit for how much the social weight can decrease when adaptive weights are enabled.

The default value of 0.5 means:

  • The attraction to the swarm's best position will never fall below this level
  • This ensures particles always maintain some connection to the group's findings

Think of it like maintaining a minimum level of communication among searchers:

  • Even when focusing on independent exploration, searchers still consider the group's best find
  • This prevents completely isolated search behavior

You might want a lower value (like 0.2) if:

  • You want more independent exploration, especially early in the search
  • You're concerned about premature convergence
  • Your problem has many distinct regions that should be explored separately

You might want a higher value (like 0.8) if:

  • You want to maintain more coordinated search behavior throughout
  • You believe early discovered promising regions should be thoroughly exploited
  • You want faster convergence toward promising solutions

This setting has no effect unless UseAdaptiveWeights is set to true.

SocialParameter

Gets or sets the social parameter that controls the influence of the swarm's global best position.

public double SocialParameter { get; set; }

Property Value

double

The social parameter, defaulting to 1.49445.

Remarks

The social parameter, also known as c2 or the social acceleration coefficient, determines how much the particle is influenced by the best position found by any particle in the swarm. This coefficient controls the "social knowledge" of the particle, encouraging it to move toward the globally best known region. Higher values increase the attraction toward the swarm's best position, promoting faster convergence but potentially at the cost of diversity. The default value of 1.49445 is derived from Clerc's constriction coefficient approach.

For Beginners: This setting controls how strongly a particle is attracted to the best position found by any particle in the entire swarm.

The default value of 1.49445 means:

  • The particle is strongly influenced by the group's collective knowledge
  • This creates a balance between exploring independently and following the group's best find

Think of it like social influence:

  • High social value: You strongly trust the group's findings and head where others have found success
  • Low social value: You're more independent and less influenced by what others have found

You might want a higher value if:

  • You want faster convergence toward promising solutions
  • You believe the global best is likely near the true optimum
  • You want the swarm to focus on refining the best known solution

You might want a lower value if:

  • You want to maintain more diversity in the search
  • You're concerned about premature convergence to local optima
  • You want more thorough exploration before committing to a region

Like the cognitive parameter, the specific default value (1.49445) is part of a mathematically derived set that provides guaranteed convergence properties.

SocialWeightAdaptationRate

Gets or sets the rate at which the social weight adapts when using adaptive weights.

public double SocialWeightAdaptationRate { get; set; }

Property Value

double

The social weight adaptation rate, defaulting to 1.0.

Remarks

This parameter controls how rapidly the social parameter changes when UseAdaptiveWeights is enabled. Higher values lead to faster adaptation, while lower values result in more gradual changes. The exact effect depends on the specific adaptive weight scheme implemented, but generally this acts as a multiplier on the amount of change applied to the social parameter in each iteration. This parameter is only relevant when UseAdaptiveWeights is set to true.

For Beginners: This setting controls how quickly the social weight (attraction to group's best position) changes over time when adaptive weights are enabled.

The default value of 1.0 means:

  • The social weight changes at a standard rate
  • Neither accelerated nor slowed down

Think of it like setting the speed for changing the level of group influence:

  • Higher values (like 1.5) make group consensus become influential more quickly
  • Lower values (like 0.5) delay the transition to strong group influence

You might want a higher value if:

  • You want faster convergence in later stages
  • You have limited iterations available for the search
  • You want more aggressive refinement once good regions are identified

You might want a lower value if:

  • You want to maintain diversity longer
  • You're concerned about premature convergence
  • You prefer more gradual transitions in search behavior

This setting has no effect unless UseAdaptiveWeights is set to true.

SwarmSize

Gets or sets the number of particles in the swarm.

public int SwarmSize { get; set; }

Property Value

int

The swarm size, defaulting to 50.

Remarks

This parameter determines the population size of the swarm. Each particle represents a candidate solution that explores the search space. Larger swarms provide better coverage of the search space and are less likely to converge prematurely to local optima, but require more computational resources. Smaller swarms converge faster but may miss the global optimum. The optimal swarm size depends on the dimensionality and complexity of the problem being solved. For many problems, a swarm size of 10-50 particles provides a good balance between exploration and computational efficiency.

For Beginners: This setting controls how many "searchers" (particles) are looking for the solution simultaneously.

The default value of 50 means:

  • 50 different particles will explore the solution space
  • Each particle moves independently but shares information with others

Think of it like a search party:

  • More people can cover more ground, but require more coordination
  • Fewer people are easier to organize, but might miss areas

You might want more particles (like 100 or 200) if:

  • Your problem is complex with many dimensions
  • You suspect there are many local optima to avoid
  • You have the computational resources to spare

You might want fewer particles (like 20 or 30) if:

  • You need faster computation
  • Your problem is relatively simple
  • You're doing preliminary exploration

Finding the right swarm size balances thorough exploration with computational efficiency.

UseAdaptiveInertia

Gets or sets whether to use adaptive inertia weight that changes throughout the optimization process.

public bool UseAdaptiveInertia { get; set; }

Property Value

bool

Whether to use adaptive inertia, defaulting to false.

Remarks

When set to true, this option enables dynamic adjustment of the inertia weight during the optimization process. Typically, a larger inertia weight is beneficial during early iterations to promote exploration, while a smaller inertia is more suitable in later iterations to refine solutions (exploitation). Adaptive inertia schemes automatically decrease the inertia weight over time or based on the swarm's performance, facilitating a natural transition from exploration to exploitation. This can improve performance on many problems compared to a fixed inertia weight.

For Beginners: This setting determines whether the inertia weight should automatically change during the optimization process.

The default value of false means:

  • The inertia weight stays constant throughout the search
  • This provides consistent, predictable particle behavior

Setting this to true means:

  • The inertia weight will typically start higher (favoring exploration)
  • And gradually decrease (shifting toward exploitation)

Think of it like a search strategy that evolves over time:

  • First phase: Broad, wide-ranging exploration with high momentum
  • Later phases: More careful, focused refinement with less momentum

You might want to set this to true if:

  • You want the algorithm to automatically shift from exploration to exploitation
  • You're dealing with complex problems that need thorough initial exploration
  • You want to avoid manually tuning the inertia parameter

You might want to keep it false if:

  • You prefer consistent, predictable behavior throughout the search
  • You've already tuned the fixed inertia weight for your specific problem
  • You're using other adaptive mechanisms in the algorithm

When set to true, the actual inertia values used will be determined by other parameters like InitialInertia, MinInertia, MaxInertia, and InertiaDecayRate.

UseAdaptiveWeights

Gets or sets whether to use adaptive cognitive and social parameters that change throughout the optimization.

public bool UseAdaptiveWeights { get; set; }

Property Value

bool

Whether to use adaptive weights, defaulting to false.

Remarks

When set to true, this option enables dynamic adjustment of the cognitive and social parameters during the optimization process. Typically, a higher cognitive parameter (personal influence) and lower social parameter (swarm influence) are beneficial in early iterations to promote diversity and exploration. In later iterations, reducing the cognitive parameter and increasing the social parameter helps the swarm converge on promising solutions. This adaptive approach can improve performance by automatically balancing exploration and exploitation throughout the search process.

For Beginners: This setting determines whether the cognitive and social parameters should automatically change during the optimization process.

The default value of false means:

  • The cognitive and social parameters remain constant throughout the search
  • Every particle maintains the same balance of personal vs. social influence

Setting this to true means:

  • Initially, particles will rely more on their personal experience (higher cognitive parameter)
  • Later, they'll be more influenced by the group's best find (higher social parameter)

Think of it like a team working on a problem:

  • Early phase: Everyone explores independently, trusting their own findings
  • Later phase: The team gradually coordinates more, focusing on the most promising areas

You might want to set this to true if:

  • You want to encourage diverse exploration early in the search
  • You want automatic convergence behavior later in the search
  • You're working on problems that benefit from this exploration-to-exploitation transition

You might want to keep it false if:

  • You want consistent, predictable behavior throughout
  • You've already tuned the fixed parameters for your specific problem
  • You prefer simpler algorithm behavior that's easier to analyze

When set to true, the actual parameter values will be determined by other settings such as InitialCognitiveWeight, InitialSocialWeight, and their min/max/adaptation rate values.

Edit this page