No abstract

No abstract

No abstract

This study introduces reverse direction supported particle swarm optimization (RDS‐PSO) with an adaptive regulation procedure. It benefits from identifying the global worst and global best particles to increase the diversity of the PSO. The velocity update equation of the original PSO was changed according to this idea. To control the impacts of the global best and global worst particles on the velocity update equation, the alpha parameter was added to the velocity update equation. Moreover, a procedure for diversity regulation based on cosine amplitude or max–min methods was introduced. Alpha value was changed adaptively with respect to this diversity measure. Besides, RDS‐PSO was implemented with both linearly increasing and decreasing inertia weight (with 1,000 and 2,000 iterations) in order to survey the effects of these variations on RDS‐PSO performances. Six most commonly used benchmark functions and three medical classification problems were selected as experimental data sets. All experimental results showed that when the grain searching ability is not so small in the last generations, the algorithm performance continues to increase. Experimental proof of it was showed up especially in RDS‐PSO using the cosine amplitude approach. Because the best results among all the RDS‐PSO types for decreasing inertia weight modes were obtained with 2,000 maximal iterations rather than 1,000 ones.

Classification success of Support Vector Machine (SVM) depends on the characteristic of given data set and some training parameters (C and σ). In literature, a few studies have been presented for regularization of these parameters which affects classification performance directly. This study proposes a new approach based on Renyi's entropy and Logistic regression methods for parameter regularization. Our regularization procedure runs at two steps. In the first step, optimal value of kernel parameter interval is found via Renyi's entropy method and optimal C value is found via logistic regression using exponential function in the next step. In addition to, this new decision support system is applied to biomedical research area via an application related to Doppler Heart Sounds (DHS). Experimental results show the efficiency of developed regularization procedure.

No abstract

An algorithm proposed using Renyi entropy clustering to improve the searching ability of traditional particle swarm optimization (PSO) is introduced in this study. Modified PSO consists of two steps. In the first step, particles in initial population are sorted according to Renyi entropy clustering method, and in the second step, some particles are removed from population and some new particles are added instead of them based on the sorted list. Thus, a reliable new initial population is created. When using sorted list from first to last with decreasing inertia weight parameter, or from last to first with increasing inertia weight parameter, a little improved search performances have been observed on three commonly used benchmark functions. However, in other two combinations of the proposed algorithm (from last to first with decreasing inertia weight and from first to last with increasing inertia weight), little worse optimization performances than traditional PSO have been noted. These four types of the proposed algorithm were run with different exchanging rate values. Thus, the representation ability of Renyi entropy clustering on initial population and the effect of organizing inertia weight parameter were evaluated together. Experimental results which were surveyed at different exchanging rate values showed the efficiency of such evaluation.

No abstract