Optimized context weighting based on the least square algorithm

Abstract

The optimized context weighting is presented. The relationship between the weighting of context models and the weighting of the description lengths corresponding to their respective context models are discussed first and it indicates that the weighting of context models is equivalent to the weighting of the their description lengths. With these discussions, the weights optimization algorithm based on the minimum description length are presented and the least square algorithm is suggested to implement the optimization of the weights. The proposed optimization algorithm is used in the compression of genome sequences. The experiment results indicate that by using the proposed weights optimization method, our context weighting-based algorithm can achieve better results than some similar algorithms reported in the literature.