A Novel Machine Learning Approach for Intelligent Spectrum Management in Cognitive Radio Networks
Kalapraveen Bagadi, Taufik Abrão, Francesco Benedetto
Abstract
This letter proposes a novel hybrid spectrum management scheme combining transfer actor-critic learning (TACT) and Q-learning algorithms to improve the cognitive radio access network’s spectrum efficiency. The TACT algorithm improves its mean opinion score over time, while the Q-learning achieves faster convergence during spectral management. Thus, this letter seeks to alleviate resource constraints by better exploiting unused communication channels. Computer simulations are carried out compared to reinforcement learning and conventional TACT algorithms. The results evidence the efficiency of our approach for intelligent spectrum management in cognitive radio networks.
Topics & Concepts
TactCognitive radioComputer scienceScheme (mathematics)Reinforcement learningConvergence (economics)CognitionArtificial intelligenceSpectrum managementSpectrum (functional analysis)Spectral efficiencyTransfer of learningMachine learningComputer networkWirelessTelecommunicationsChannel (broadcasting)PsychologyMathematicsMathematical analysisQuantum mechanicsPhysicsPsychotherapistNeuroscienceEconomicsEconomic growthCognitive Radio Networks and Spectrum SensingAdvanced Adaptive Filtering TechniquesAdvanced MIMO Systems Optimization