Fixed budget quantized kernel least-mean-square algorithm

• Published in: Signal processing Vol. 93; no. 9; pp. 2759 - 2770
• Main Authors: Zhao, Songlin, Chen, Badong, Zhu, Pingping, Príncipe, José C.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.09.2013; Elsevier
• Subjects: Algorithms; Applied sciences; Budgeting; Detection, estimation, filtering, equalization, prediction; Dictionaries; Exact sciences and technology; Fixed budget; Growing and pruning; Information, signal and communications theory; Kernel methods; Kernels; Methodology; On-line systems; Online; Preserves; Quantized kernel least mean square; Signal and communications theory; Signal, noise; Telecommunications and information theory; Kernel methods; Quantized kernel least mean square; Fixed budget; Growing and pruning; Dictionaries; Pruning algorithm; Updating; Computational complexity; Kernel method; Accuracy; On line processing; A priori estimation; Recursive method; Signal processing; Least mean squares methods
• ISSN: 0165-1684; 1872-7557
• DOI: 10.1016/j.sigpro.2013.02.012

This paper presents a quantized kernel least mean square algorithm with a fixed memory budget, named QKLMS-FB. In order to deal with the growing support inherent in online kernel methods, the proposed algorithm utilizes a pruning criterion, called significance measure, based on a weighted contribution of the existing data centers. The basic idea of the proposed methodology is to discard the center with the smallest influence on the whole system, when a new sample is included in the dictionary. The significance measure can be updated recursively at each step which is suitable for online operation. Furthermore, the proposed methodology does not need any a priori knowledge about the data and its computational complexity is linear with the center number. Experiments show that the proposed algorithm successfully prunes the least “significant” centers and preserves the important ones, resulting in a compact KLMS model with little loss in accuracy. ► An efficient pruning and growing strategy in designing fixed-budget QKLMS is proposed. ► Priori knowledge of the input distribution is not required. ► Recursive calculation leads to a updating strategy to be computed in real time.