Skip to main content

Table 1 Learning techniques applications in CR, strengths and limitations

From: Recent advances on artificial intelligence and learning techniques in cognitive radio networks

Learning technique

Spectrum sensing (SS)

Decision-making

Strengths

Limitations and challenges

   

Adaptation ability to minor changes

Require training data labels

Neural networks

×

×

Construction using few examples,

Poor generalization

   

thus reducing the complexity

Overfitting

Support

  

Generalization ability

Requires training data labels

vector

×

×

Robustness against noise and outliers

and previous knowledge of the system

machine

   

Complex with large problems

   

Multi-objective optimization

Require prior knowledge of the system

Genetic algorithms

 

×

Dynamically configure the CR

Suitable fitness function

   

based on environment changes

High complexity with large problems

Game theory

Related to the capabilities of the

×

Reduces the complexity of adaptation

Requires prior knowledge of the system

 

spectrum-sensing technique used

 

Solutions for multi-agent systems

and labeled training data

Reinforcement

×

×

Learning autonomously using feedback

Needs learning phase of the policies

learning

  

Self-adaptation progressively in real time

 

Fuzzy logic

Related to the capabilities of the

×

Simplicity, decisions are

Needs rules derivation

 

spectrum-sensing technique used

 

directly inferred from rules

Accuracy is based on these rules

Entropy approach

×

×

Statistical model

Requires prior knowledge of the system

 

Related to the capabilities of the

 

Simplicity

Requires prior knowledge of the system

Decision tree

spectrum-sensing technique used

×

Decision using tree branches

May suffer overfitting

    

Requires labeled training data

Artificial

 

×

Parallel search for solutions

Requires prior knowledge of the system

bee colony

   

Requires a fitness function

Bayesian

×

×

Probabilistic models

Requires prior knowledge of the system

    

May face computational complexity

Markov model

×

×

Statistical models

Requires prior knowledge of the system

   

Scalable

May face computational complexity

Case-

  

Find acceptable solution based

Complex search in large databases

based

 

×

on the existing case found

Requires predefined and relevant cases

reasoning

  

in the case database

Mistakes propagation