Author Affiliations
Key Laboratory of Electromagnetic Wave Information Science, Ministry of Education, Department of Communication Science and Engineering, Fudan University, Shanghai 200433, Chinashow less
Fig. 1. Block diagram of machine learning based VLC system
Fig. 2. Application of machine learning in visible light communication
Fig. 3. Diagram of post equalization of K-means algorithm, in which black points are receiving constellation points after post-equalization, and I and Q represent in-phase component and orthogonal component of receiving data, respectively. (a) Receiving constellation and normal decision board, in which points in black circle will be misjudged; (b) CAPD decision board
Fig. 4. Flowchart of K-means algorithm
Fig. 5. Principle diagram of K-means algorithm based pre-equalization
Fig. 6. Diagrams of PAM4 signal fluctuation and DBSCAN re-classification
[33]. (a) Fluctuation of PAM4 receiving signal; (b) diagram of DBSCAN reclassification
Fig. 7. Description of core points, accessory points, and noise points of DBSCAN
Fig. 8. Description of SVM classification
Fig. 9. Effects of SVM classification and phase correction. (a) Receiving constellation before phase correction; (b) effect of SVM classification, in which Four colors at the bottom represent the four categories of the four constellation points according to the QPSK data, and Red, green, blue colors represent the input training set; (c) effect after phase correction
Fig. 10. Neural network structure of GK-DNN
Fig. 11. Structure of ANN
ML algorithm | Application | Actionposition | Supervision | Modulationformat | Generalization | Computationcomplexity |
---|
K-means | Pre-equPost-equ | TxRx | No | CAP-16QAM | Weak | Low | DBSCAN | JitterMitigation | Rx | No | CAP-16QAM | Weak | Low | SVM | Phaseestimation | Rx | Yes | CAP-QPSK | Middle | Middle | ANN | Post-equ | Rx | Yes | PAM,QAM | Strong | Middle | GK-DNN | Nonlinearmitigation | Rx | Yes | PAM8 | Strong | High |
|
Table 1. Summarization of machine learning algorithms