Optimization of Machine Learning Parameters for Spectrum Survey Analysis
This paper shows preliminary results of the optimization of machine learning parameters for cognitive radio application by brutal force calculations.We were analyzing frequency occupancy data of the huge measurement campaign of the spectrum background.For these date there are two possible states.Firstly,limited frequency band is occupied(detected signal level is above the threshold)by the other frequency signal-there will be an interference for our system for this frequency band.Secondly,the frequency band is free of any other wireless radiation.These true/false data are analyzed in a context of the cognitive radio by the reinforcement learning and simple learning.Each channel received a score from the learning algorithm given by weighting function.The quality of the output scores is discussed in this paper according to the learning algorithm parameters and optional learning time.
R.Urban M.Steinbauer
Department of Theoretical and Experimental Electrical Engineering Brno University of Technology,Technicka 12,Brno 612 00,Czech Republic
国际会议
Progress in Electromagnetics Research Symposium 2014(2014年电磁学研究新进展学术研讨会)
广州
英文
616-619
2014-08-01（万方平台首次上网日期，不代表论文的发表时间）