PaperlayerView exact revision on arXiv

Source-linked AI summary

Genetic Algorithms in Wireless Networking: Techniques, Applications, and Issues

Usama Mehboob, Junaid Qadir, Salman Ali, Athanasios Vasilakos

arXiv:1411.5323v1cs.NI

TL;DR

Wireless networking’s dynamic and often unknown conditions make parameter optimization challenging. This paper surveys GA models, configurations, and wireless applications, alongside implementation challenges and open issues. It concludes with a categorized account of applications and future research directions.

  • Problem

    Dynamic and often unknown wireless-network conditions make parameter optimization challenging.

    Source: 2) Adaptiveness and Online Problem… · p. 2

  • Method

    The paper surveys GA models and configurations and categorizes wireless-network applications by network configuration and GA technique.

    Source: §VI · p. 23

  • Results

    The survey covers wireless-network GA applications, common configurations, implementation pitfalls, and open issues, and identifies directions for future work.

    Source: §VI · p. 23

  • Takeaways & Limitations

    The paper’s synthesis provides a categorized account of GA applications and highlights implementation challenges and open research directions in wireless networking.

    Source: §VI · p. 23

    GAs lack mathematical guarantees of finding a global optimum and can stall near a suboptimal point when population diversity or size is inadequate.

    Source: 1) Genetic Drift or Bias · p. 21

Abstract

from arXiv · show

In recent times, wireless access technology is becoming increasingly commonplace due to the ease of operation and installation of untethered wireless media. The design of wireless networking is challenging due to the highly dynamic environmental condition that makes parameter optimization a complex task. Due to the dynamic, and often unknown, operating conditions, modern wireless networking standards increasingly rely on machine learning and artificial intelligence algorithms. Genetic algorithms (GAs) provide a well-established framework for implementing artificial intelligence tasks such as classification, learning, and optimization. GAs are well-known for their remarkable generality and versatility, and have been applied in a wide variety of settings in wireless networks. In this paper, we provide a comprehensive survey of the applications of GAs in wireless networks. We provide both an exposition of common GA models and configuration and provide a broad ranging survey of GA techniques in wireless networks. We also point out open research issues and define potential future work. While various surveys on GAs exist in literature, our paper is the first paper, to the best of our knowledge, which focuses on their application in wireless networks.

I. INTRODUCTION

Genetic algorithms draw on biological evolution and use encoded candidate solutions with mutation, crossover, and selection. Their population-based search is presented as adaptable to wireless networks’ changing conditions and diverse optimization needs.

1) Generality and Versatility:

  • GAs are general metaheuristics for large search spaces and can evaluate solutions without explicitly knowing the objective function, but they do not guarantee optimality.Mutation and crossover can help avoid local optima with high probability.Source: 1) Generality and Versatility · p. 1

2) Adaptiveness and Online Problem Solving:

  • In changing wireless conditions, GAs can provide satisfactory adaptive performance as network nodes interact and their decisions depend on one another.The paper contrasts these dancing landscapes with traditional optimization assumptions of static, well-known topologies.Source: 2) Adaptiveness and Online Problem… · p. 2
  • Online GAs balance exploration of unfamiliar solutions with exploitation of known good ones; mutation and recombination correspond to exploration, while selection corresponds to exploitation.The paper says continued exploration matters in ever-changing environments; near-optimal balance under Holland’s schema analysis is qualified by assumptions.Source: 2) Adaptiveness and Online Problem… · p. 2

3) Ability to Find Good Building Blocks:

  • GAs can combine high-performing chromosome substrings, called schemas, to form stronger solutions.The passage describes these substrings as building blocks of the overall solution.Source: 3) Ability to Find Good Building Blocks · p. 2

4) Parallel Nature and Scalability:

  • GAs’ population-based evolutionary operations provide a parallel framework for scalable problem solving across applications.The described operations include mutation, crossover, selection, and recombination.Source: 4) Parallel Nature and Scalability · p. 2

5) Support For Multiobjective Optimization:

  • GAs support joint optimization of multiple potentially conflicting objectives, a practical need in wireless networks.The paper gives signal-to-interference-and-noise ratio, power consumption, and bit error rate as wireless objectives.Source: 5) Support For Multiobjective… · p. 2

6) Support for Global Optimization:

  • GAs search probabilistically over encoded parameters with a population and a fitness function that does not require derivatives, properties the paper associates with global optimization.The section contrasts GAs with multilayer perceptrons, which it describes as making local changes to find local minima.Source: 6) Support for Global Optimization · p. 2

7) Easy Implementation:

The paper surveys GA applications in wireless networking, introduces common GA concepts, and outlines techniques and open research directions. It also describes GAs as computationally simple and readily implementable.

7) Easy Implementation:

  • GAs are described as computationally simpler than neural networks and suitable for rapid prototyping on DSPs or FPGAs and implementation on parallel and cloud architectures.The paper attributes their parallel-computing suitability to their inherently parallel nature.Source: 7) Easy Implementation · p. 2

B. Contributions of this Paper

  • The survey covers GA applications across wireless mesh, mobile ad-hoc, cellular, sensor, and cognitive radio networks, while aiming for a representative sample rather than exhaustive coverage.The paper says the breadth of GA literature precludes an exhaustive survey.Source: B. Contributions of this Paper · pp. 2–3

C. Organization of this Paper

  • The paper organizes its coverage around GA background, wireless applications, implementation lessons, and open issues and future directions.Listed application tasks include routing, channel allocation, load balancing, localization, and quality of service.Source: C. Organization of this Paper · p. 3

II. BACKGROUND: GENETIC ALGORITHMS

A. GA Terminology

  • In the paper’s terminology, an organism is the entity being optimized, a population is the collection undergoing evolution, and a chromosome encodes a solution.A radio parameter or wireless resource can be an organism; chromosomes contain genes, which define features and take alternative forms called alleles.Source: A. GA Terminology · p. 3
  • Fitness denotes a chromosome’s utility, selection retains fitter simulated organisms, mutation changes an allele, and locus names a gene’s chromosome position.The terminology frames GA operations using biological metaphors.Source: A. GA Terminology · p. 3

B. Number-of-Objectives Based Classification

The section distinguishes GA configurations by objective count and by adaptation or parallelism. It describes tradeoffs in repeated single-objective optimization, Pareto compromises, parameter adaptation, and parallel architectures.

1) Single-Objective GAs:

  • Single-objective GAs optimize a scalar fitness function and are less computationally exhaustive, but multiple objectives may require repeated runs with different weight vectors.The passage contrasts this with multi-objective GAs.Source: 1) Single-Objective GAs · pp. 3–4

2) Multi-objective GAs:

  • Multi-objective GAs seek non-dominated solutions on a Pareto front, where improving one objective requires degrading another, so solutions represent compromises.The examples include minimizing power versus minimizing bit error rate.Source: 2) Multi-objective GAs · p. 4

1) Adaptive GA:

  • Adaptive GAs use information about each generation’s population to adjust crossover and mutation probabilities, unlike simple GAs with fixed control parameters.The section presents adaptation as a technique to prevent suboptimal convergence to local optima.Source: 1) Adaptive GA · p. 5

2) Parallel GA:

  • Parallel GAs include master-slave, island, and cellular forms; island GAs are described as best suited to wireless networking because they allow control of migration policy.The passage contrasts this with communication demands in cellular and master-slave configurations.Source: 2) Parallel GA · p. 5

3) Distributed or Island GA:

GAs evolve populations of candidate solutions through selection and genetic operators, while distributed island models divide populations into cooperating subpopulations. Encoding choices represent problem-specific solutions as chromosomes, including variable-length representations for routing paths.

3) Distributed or Island GA:

  • Island GAs divide populations into subpopulations that exchange individuals under a migration policy, with relatively infrequent migration reducing inter-agent communication requirements.Each island runs a GA locally, and migration governs when and how individuals move between islands.Source: 3) Distributed or Island GA · p. 5

D. GA Components And Operators

  • Fitter individuals are more likely to be selected and mutated or recombined into the next generation, gradually improving fitness and helping GAs avoid local optima.Populations may start randomly, but an appropriately defined initial population can accelerate convergence.Source: D. GA Components And Operators · p. 5

1) Encoding into Chromosomes:

  • Chromosomes can use binary, permutation, value, or tree encodings, with the representation matched to the problem; permutation encoding suits ordering problems but may require repair after operators.Value encoding accommodates complicated values but often requires problem-specific mutation and crossover operators; tree encoding represents expressions or programs.Source: 1) Encoding into Chromosomes · pp. 5–6
  • Variable-length chromosomes can represent routing paths of different lengths, with path genes encoding node IDs and chromosome length bounded by the network's node count N.A radio-parameter encoding proposal uses variable bit lengths because parameters such as frequency and modulation require very different numbers of bits.Source: 1) Encoding into Chromosomes · pp. 5–6

2) Fitness Function:

Fitness functions score candidate solutions against optimization goals and constraints, but designing them can require expert effort and fast computation. Mutation alters chromosome genes to explore alternatives and diversify the population.

2) Fitness Function:

  • A fitness function evaluates chromosomes against optimization goals and constraints, assigning scores that guide which designs are retained or replaced.For routing, fitness may represent high average throughput or low delay; classification fitness may measure error.Source: 2) Fitness Function · pp. 6–7
  • Fitness-function design requires human expertise, a precise model, low computation time, and noise control; an inefficient function can yield inappropriate solutions or no convergence.Repeated evaluation makes execution speed important, while some problems make the fitness definition difficult or impossible to specify, motivating interactive evaluation.Source: 2) Fitness Function · p. 7

3) Mutation:

  • Mutation randomly alters chromosome genes to explore new traits and diversify the population, helping GAs avoid local optima.Non-uniform mutation can reduce mutation extent over generations and is applicable to float and integer genes.Source: 3) Mutation · p. 7

4) Crossover:

Crossover recombines parent chromosomes to exploit their traits, while selection favors fitter individuals for reproduction. The section describes crossover variants and several ways to select or preserve candidates.

4) Crossover:

  • Crossover exchanges chromosome subsequences between two parents to create offspring, combining existing traits; together with mutation, it supplies small-step and exploratory search.Crossover points may be multiple to increase exploration.Source: 4) Crossover · p. 7
  • One-point and two-point crossover swap segments, whereas uniform crossover mixes genes across the chromosome and half-uniform crossover swaps half of the nonmatching bits.Cut-and-splice crossover allows offspring strings to change length because each parent has its own crossover point.Source: 4) Crossover · p. 8

5) Selection:

  • Selection chooses individuals for reproduction with fitter chromosomes more likely to advance; roulette-wheel selection makes selection probability proportional to fitness.Tournament selection repeatedly selects the best from a random subset, while elitism preserves previously found best solutions across generations.Source: 5) Selection · pp. 8–9
  • Relative pooling tournament evaluation compares each parent against population members, but its computational overhead can stress GA deployment.For multi-objective optimization, archive maintenance checks new solutions for non-dominance, favoring a modest bounded archive size.Source: 5) Selection · p. 9

III. APPLICATIONS OF GAS IN WIRELESS NETWORKING

The survey organizes wireless-network GA applications by task and network type, and also covers hybrid approaches. Its examples include routing and QoS methods, with reported benefits for dynamic routing and tree-based QoS composition.

A. Application Based Classification

1) Routing:

  • For shortest-path routing, predefined candidate routes can be evolved using variable-length chromosomes, with infeasible solutions checked during optimization.The number of possible routes can become too large to enumerate in dense or large networks.Source: 1) Routing · p. 9
  • Immigrant and memory schemes enhance GA performance for dynamic shortest-path routing in MANETs.The schemes were adapted and integrated into a specialized GA for the dynamic shortest-path routing problem.Source: 1) Routing · pp. 9–10
  • A hybrid ant-algorithm and GA approach was reported to generate routes and achieve efficient load balancing.The ant algorithm reduces routing-table size, while the combined techniques independently explore the network.Source: 1) Routing · pp. 10–11

2) Quality of Service (QoS):

  • GA-based QoS work includes near-optimal call admission control for large wireless/mobile networks and multi-objective routing that can optimize user-prioritized requirements.The call-admission approach offers adaptive QoS to multi-class users and is computationally less intensive than the SMDP approach.Source: 2) Quality of Service (QoS) · p. 11
  • Tree-based GA models for QoS composition have less overhead than traditional serial-chromosome GAs while supporting replanning and service composition.Subtree crossover exchanges task nodes and updates the QoS vector from leaf nodes toward the tree root.Source: 2) Quality of Service (QoS) · p. 11

3) Load Balancing:

The survey reviews GA approaches to load balancing, bandwidth allocation, and channel assignment, including how they adapt wireless resources to congestion, QoS needs, and interference constraints.

3) Load Balancing:

  • MicroGA initially outperforms MacroGA, but MacroGA later improves load balancing by moving more congested users toward peripheral access points.The surveyed work also describes RSSI, least-loaded-first, and hybrid load-balancing categories, and considers QoS weighting by user service requirements.Source: 3) Load Balancing · p. 11

4) Bandwidth Allocation:

  • An accelerated GA varies mutation and crossover probabilities with chromosome fitness to reduce generations and computation time while converging quickly to a global optimum.Another approach assigns multimedia substreams among high, medium, and low QoS levels, degrading some admitted calls under congestion to preserve resources and promote fair bandwidth use.Source: 4) Bandwidth Allocation · p. 13

5) Channel Allocation and Assignment:

  • Channel assignment must account for co-channel, adjacent-channel, and co-site interference constraints.The section frames channel allocation as maximizing frequency reuse while minimizing the number of channels needed to cover network cells.Source: 5) Channel Allocation and Assignment · pp. 13–14
  • >14.5 dB SNR gain and higher throughput were achieved with GAs versus least-interference and channel-segregation methods for dynamic channel assignment.The reported comparison is for a distributed channel-assignment approach in broadband fixed wireless access networks.Source: 5) Channel Allocation and Assignment · p. 14
  • A hybrid GA with a local-improvement operator targets more connections while minimizing isochronous channels in multiple-access telecommunication networks.For cellular networks, the survey also reports GA use in cluster sizing and notes that increasing cells per cluster slows convergence and increases generations.Source: 5) Channel Allocation and Assignment · p. 14
  • The genetic-fix algorithm uses fixed-size individuals and minimum-separation encoding to reduce the search space and the bits needed to represent solutions.The surveyed work distinguishes fixed channel assignment, where slots are pre-allocated to cells, from dynamic assignment upon request.Source: 5) Channel Allocation and Assignment · p. 14

6) Localization:

The localization survey covers GA-based node positioning and placement across wireless settings, with examples using anchor nodes, hybrid search, and optimization of coverage or connectivity.

Localization

  • Other surveyed applications include aeronautical gateway allocation with TDMA scheduling to minimize latency, and MANET node placement for mesh-network connectivity and user coverage.The examples also include GA-assisted decisions about node mobility and organization, and adjusting mobile nodes’ speed, location, and direction over terrain.Source: 6) Localization · p. 15
  • GA-based placement work targets WSN coverage and mesh connectivity, using approaches including GP-assisted deployment and grid-based node placement.The supplied examples also describe hybrid GA and greedy optimization of node placement for traffic balancing in two-tier heterogeneous WSNs, and placement in obstacle-filled grids for real-time settings.Source: 6) Localization · p. 15
  • Tam et al. locate each node relative to three anchor nodes by triangulation, then use a centralized GA as a post-optimizer for the network graph.Zhang et al. combine a centralized GA with simulated annealing to minimize WSN position-estimation error; simulated annealing is described as a local-search technique that avoids premature convergence.Source: 6) Localization · p. 15

7) WLAN AP placement:

The survey reviews GA applications to WLAN and cellular access-point placement, then describes additional uses in network planning and forecasting.

7) WLAN AP placement:

  • The MOGA approach produces solutions with a high-SNR data profile, with more network area points having average SNR.The placement objectives are to minimize the number of access points and maximize SNR.Source: 7) WLAN AP placement · pp. 15–16
  • Indoor base-station placement uses a microwave propagation model that accounts for floors, ceilings, walls, and obstacles such as brick, concrete, wood, furniture, and glass.The surveyed approach assigns obstacle-specific weights and uses a parallel GA heuristic to optimize indoor coverage.Source: 7) WLAN AP placement · p. 16
  • For cellular and mobile coverage, a GA selects base-station sets against power and SNR constraints, and its fitness function can incorporate other network requirements.The described approach combines simulation with evolutionary techniques and defines coverage areas using received power greater than -60dbm.Source: 7) WLAN AP placement · p. 16

8) General Applications:

  • General applications include GA-based base-station sleep-mode planning and optimization of node distribution, antenna patterns, and SNR.The surveyed antenna approach uses binary-coded candidate distributions and gives greater inheritance probability to chromosomes with minimum path-length cost.Source: 8) General Applications · p. 17
  • GA-based forecasting estimates product sales and peak timing from prior data and factors such as price, competition, and network effects; a hybrid GA-neural-network approach forecasts cellular churn.The cited churn work reports greater accuracy than statistical models but higher computational expense.Source: 8) General Applications · p. 17

B. Network Type Based Classification

The survey organizes GA applications by wireless-network type, including WSNs, WMNs, and CRNs, and describes their use in coverage, placement, and radio adaptation.

1) Wireless Sensor Networks (WSN):

  • WSN applications target area coverage with fewer nodes while preserving energy, using multi-objective coverage control and adaptive choices of active sensors and clusterheads.Other surveyed work forms clusters to reduce communication distances; keeping clusters at 10% of total nodes is reported to reduce path cost and communication distance by 80% versus random distribution.Source: 1) Wireless Sensor Networks (WSN) · p. 17
  • A GA using anchor-node distances reports lower mean position error than SDPL and SAL for WSN localization.One-hop distance estimates are collected by clusterheads to construct the network graph before GA-based location estimation.Source: 1) Wireless Sensor Networks (WSN) · pp. 17–18

2) Wireless Mesh Networks (WMNs):

  • For WMNs, GA placement of mesh routers targets user coverage and connectivity, and the surveyed approach achieves connectivity across client distributions while coverage depends on client placement.The approach encodes clients and routers on rectangular grids and uses crossover and mutation to alter router placements.Source: 2) Wireless Mesh Networks (WMNs) · p. 18

3) Cognitive Radio Networks (CRNs):

  • CRN multi-objective GAs optimize time-varying QoS across parameters such as BER, throughput, power consumption, and interference as radio conditions and spectrum availability vary.The survey describes using remembered prior objectives and winner chromosomes to partially initialize a new GA population, while random individuals preserve diversity.Source: 3) Cognitive Radio Networks (CRNs) · pp. 18–19
  • CRN cognitive engines use channel information and learning to update radio configurations, with WCGA, CSM, and WSGA handling channel modeling, learning and goals, and GA-based configuration search.GA chromosomes encode radio traits such as modulation, frequency, spreading, filtering, and FEC coding; objectives include BER, power consumption, and interference.Source: 3) Cognitive Radio Networks (CRNs) · pp. 18–19
  • Cognitive-radio testbeds apply GA operators across radio layers, and a separate two-step GA testbed optimizes channel selection and transmitter parameters in separate iterations.Simulation of the latter is reported to converge bandwidth and transmit power to optimal values within a few GA iterations.Source: 3) Cognitive Radio Networks (CRNs) · p. 19, Figure 7 · p. 19

4) Cellular Networks (CNs):

The survey describes GA applications in cellular and software-defined wireless networks, then reviews hybrids combining GAs with other techniques. Examples span resource allocation, channel assignment, and several forms of network optimization.

4) Cellular Networks (CNs):

  • Cellular-network GAs address resource allocation and energy efficiency, including adaptive spectral-efficiency optimization and base-station sleep-mode deployment.The cited work adjusts resources between relay and base stations and uses sleep modes to reduce transmitter power and count.Source: 4) Cellular Networks (CNs) · p. 19
  • Immune operators, compatibility matrices, and adaptive mutation have been used to pursue conflict-free channel assignment in cellular networks.Related work also applies adaptive GAs to dynamic and fixed assignments and uses a micro GA for fixed assignment.Source: 4) Cellular Networks (CNs) · pp. 19–20
  • Adaptive GAs for cellular dynamic spectrum access track channel capacity and existing calls while exploring channel choices to address call dropping and blocking.The approaches use random immigrants and parameter adaptation.Source: 4) Cellular Networks (CNs) · p. 20

5) Software Defined Wireless Networks (SWNs):

  • In software-defined wireless networks, a GA-based small-cell mobile-backhaul resource manager has been proposed to work with software-defined management.The passage characterizes this networking configuration as relatively new.Source: 5) Software Defined Wireless Networks… · p. 20

C. Hybridization: GA With Other Techniques

  • The survey frames hybridization as combining GAs with other adaptive methods, including game theory, reinforcement learning, graph theory, fuzzy logic, and quantum approaches.It presents combining methods as a way to use different techniques as building blocks.Source: C. Hybridization: GA With Other… · p. 20

1) Combining GAs with Game Theory:

  • Evolutionary game theory applies evolutionary computing and GAs to network interactions including sensing, access and power control, routing, selection, congestion control, TCP, and bandwidth allocation.The cited passage lists these as contexts for evolutionary games in networks.Source: 1) Combining GAs with Game Theory · p. 20

2) Combining GAs with Reinforcement Learning:

  • GAs are described as policy-space reinforcement learning: mutation and recombination explore policies, while selection retains stronger ones.Policies are encoded in chromosomes and evaluated by cumulative fitness.Source: 2) Combining GAs with Reinforcement… · p. 20

3) Combining GAs with Graph Theory:

  • Evolutionary graph methods use evolving graphs to model MANET connectivity and can estimate congestion and suggest parallel routes using foremost journeys.A separate hybrid approach combines graph theory with evolutionary techniques for decentralized, energy-conserving forwarding.Source: 3) Combining GAs with Graph Theory · p. 20

5) Combining GAs with Quantum Theory:

The section surveys hybrid GA variants and practical issues, including search performance, convergence, and implementation choices. It cautions that outcomes depend on problem structure and GA configuration.

5) Combining GAs with Quantum Theory:

  • Quantum GAs use qubits and rotation gates to preserve diversity and are reported more effective than traditional GAs for larger inputs; a comparison also reports faster, more robust performance than classical GAs on the traveling salesperson problem.The larger-input comparison is attributed to linear superposition states.Source: 5) Combining GAs with Quantum Theory · p. 21

6) Combining GAs with Local Search:

  • Memetic algorithms combine GAs with local search and domain-specific knowledge, potentially strengthening results at higher computational cost.The local search operator is applied to each population member after every generation.Source: 6) Combining GAs with Local Search · p. 21

7) Combining GAs with Other Metaheuristics:

  • GA hybrids with methods such as ant colony optimization, neural networks, and fuzzy logic target dynamic constrained multi-objective optimization, with further improvement still possible.The passage specifically notes work adjusting neural-network connection weights and connectivity.Source: 7) Combining GAs with Other… · p. 21

1) Genetic Drift or Bias:

  • GA search can converge slowly or settle at a local minimum when parameters are poorly chosen or population diversity is inadequate, and there is no guarantee of a global optimum.The passage also identifies poor fitness functions as a source of poor chromosome blocks.Source: 1) Genetic Drift or Bias · p. 21

2) Using GAs for Real-Time Applications:

  • Stochasticity and heuristic search make GAs unsuitable for real-time applications requiring guaranteed response times, while wireless QoS constraints may force termination after a preset iteration count.Response time can vary with control parameters and population initialization.Source: 2) Using GAs for Real-Time Applications · p. 21

3) GA Deception:

  • GA-deceptive objective functions can make good building blocks combine into poor chromosomes; messy genetic algorithms are offered as a potential remedy.The passage does not present the remedy as a guarantee.Source: 3) GA Deception · p. 21

1) Setting Population Size:

  • Population size affects solution quality and convergence speed; for moderately complex problems, 50–100 chromosomes may be a default, but the choice depends on complexity and search variables.Population size may also grow or contract based on convergence and genetic diversity.Source: 1) Setting Population Size · pp. 21–22

2) Designing Solution Encoding and Fitness Function:

  • Chromosome encoding and fitness-function design are critical to GA efficiency and convergence, with the appropriate representation depending on the problem rather than a one-size-fits-all rule.Bit representation is common, but data structures may suit some problems better.Source: 2) Designing Solution Encoding and… · p. 22

3) Setting of Mutation and Crossover Rates:

The section offers implementation guidance and identifies open research directions, from theoretical understanding and problem-specific knowledge to distributed optimization and scaling. It concludes by recapping the survey’s coverage of applications, techniques, pitfalls, and future work.

3) Setting of Mutation and Crossover Rates:

  • For large populations of about 100, crossover and mutation rates of 0.6 and 0.001 are proposed, while small populations of about 30 use 0.9 and 0.01.Another recommendation is mutation probability 0.05 and crossover rate 0.6 with a population of about 50.Source: 3) Setting of Mutation and Crossover… · p. 22

4) Computation of Fitness Values:

  • When fitness computation is computationally intensive, approximation can avoid executing the real-world system to evaluate candidate solutions.The passage identifies modeling large complex systems as a practical challenge.Source: 4) Computation of Fitness Values · p. 22

5) Setting of Termination Criteria:

  • Termination criteria should define a context-dependent good-enough solution; 2.5 times the population size is offered as a reasonable maximum-generation estimate.The passage says the criterion depends on the number of objectives and variables.Source: 5) Setting of Termination Criteria · p. 22

A. Theoretical Results for GAs

  • The authors call for stronger GA theory, noting slow progress in rigorous results, the No Free Lunch implication, and no guarantee that each individual in an improved population improves.The No Free Lunch principle means algorithms perform equally when averaged across all problems, so comparative advantages are problem-class-specific.Source: A. Theoretical Results for GAs · p. 22

B. How to Exploit Problem Specific Knowledge?

  • Embedding problem-specific knowledge may accelerate GA search, but specialized heuristics or operators can reduce generality; wireless-specific knowledge remains an open issue.The passage discusses this as a question requiring further exploration.Source: B. How to Exploit Problem Specific… · p. 22

C. Efficient Multi-Objective and Distributed GAs; D. GAs, Complexity Theory, and Wireless Networks; E. Using Big Data Techniques For Scaling GAs

VI. CONCLUSIONS

  • The conclusion summarizes a survey of GA applications organized by wireless-network configuration and GA technique, alongside implementation pitfalls and future research directions.The paper also provided an introduction to common GA models and configurations.Source: §VI · p. 23
Loading 1411.5323v1…