• Journal of Institute of Control, Robotics and Systems
• /
• v.17 no.12
• /
• pp.1210-1218
• /
• 2011

GA (Genetic Algorithms) are efficient for searching for global optima but may have some problems such as premature convergence, convergence to local extremum and divergence. These phenomena are related to the evolutionary operators. As population diversity converges to low value, the search ability of a GA decreases and premature convergence or converging to local extremum may occur but population diversity converges to high value, then genetic algorithm may diverge. To guarantee that genetic algorithms converge to the global optima, the genetic operators should be chosen properly. In this paper, we analyze the effects of the selection operator, crossover operator, and mutation operator on convergence properties, and propose the sweeping method of mutation probability and elitist propagation rate to maintain the diversity of the GA's population for getting out of the premature convergence. Results of simulation studies verify the feasibility of using these sweeping operators to avoid premature convergence and convergence to local extrema.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.5 no.1
• /
• pp.21-28
• /
• 2005

This paper proposes a new method for accelerating the search speed of genetic algorithms by taking derivative evaluation and conditional random selection into account in their evolution process. Derivative evaluation makes genetic algorithms focus on the individuals whose fitness is rapidly increased. This accelerates the search speed of genetic algorithms by enhancing exploitation like steepest descent methods but also increases the possibility of a premature convergence that means most individuals after a few generations approach to local optima. On the other hand, derivative evaluation under a premature convergence helps genetic algorithms escape the local optima by enhancing exploration. If GAs fall into a premature convergence, random selection is used in order to help escaping local optimum, but its effects are not large. We experimented our method with one combinatorial problem and five complex function optimization problems. Experimental results showed that our method was superior to the simple genetic algorithm especially when the search space is large.

• Journal of Advanced Navigation Technology
• /
• v.18 no.2
• /
• pp.170-177
• /
• 2014

Genetic Algorithms are robust search and optimization techniques but have some problems such as premature convergence and convergence to local extremum. As population diversity converges to low value, the search ability decreases and converges to local extremum but population diversity converges to high value, then the search ability increases and converges to global optimum or genetic algorithm may diverge. To guarantee that genetic algorithms converge to the global optima, the genetic operators should be chosen properly. In this paper, we propose the genetic operators with the hybrid function of the average Hamming distance and the fitness value to maintain the diversity of the GA's population for escaping from the premature convergence. Results of simulation studies verified the effects of the mutation operator for maintaining diversity and the other operators for improving convergence properties as well as the feasibility of using proposed genetic operators on convergence properties to avoid premature convergence and convergence to local extremum.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.10 no.4
• /
• pp.263-268
• /
• 2010

The premature convergence of genetic algorithms (GAs) is the most major factor of slow evolution of GAs. In this paper we propose a novel method to solve this problem through competition of multiple offspring of in dividuals. Unlike existing methods, each parents in our method generates multiple offspring and then generated multiple offspring compete each other, finally winner offspring become to real offspring. From this multiple offspring competition, our GA rarel falls into the premature convergence and easily gets out of the local optimum areas without negative effects. This makes our GA fast evolve to the global optimum. Experimental results with four function optimization problems showed that our method was superior to the original GA and had similar performances to the best ones of queen-bee GA with best parameters.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.34S no.8
• /
• pp.92-103
• /
• 1997

This paper proposes a new gray-level image segmentation method using GA(genetic algorithm) and an ACA(adaptive clustering algorithm). The solution in the general GA can be moving because of stochastic reinsertion, and suffer from the premature convergence problem owing to deficiency of individuals before finding the optimal solution. To cope with these problems and to reduce processing time, we propose the new GBR algorithm and the technique that resolves the premature convergence problem. GBR selects the individual in the child pool that has the fitness value superior to that of the individual in the parents pool. We resolvethe premature convergence problem with producing the mutation in the parents population, and propose the new method that removes the small regions in the segmented results. The experimental results show that the proposed segmentation algorithm gives better perfodrmance than the ACA ones in Gaussian noise environments.

• The KIPS Transactions:PartB
• /
• v.17B no.2
• /
• pp.149-156
• /
• 2010

Since the premature convergence phenomenon of genetic algorithms (GAs) degrades the performances of GAs significantly, solving this problem provides a lot of effects to the performances of GAs. In this paper, we propose a selective mutation method in order to improve the performances of GAs by alleviating this phenomenon. In the selective mutation, individuals are additionally mutated at the specific region according to their ranks. From this selective mutation, individuals with low ranks are changed a lot and those with high ranks are changed small in the phenotype. Finally, some good individuals search around them in detail and the other individuals have more chances to search new areas. This results in enhancing the performances of GAs through alleviating of the premature convergence phenomenon. We measured the performances of our method with four typical function optimization problems. It was found from experiments that our proposed method considerably improved the performances of GAs.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.10 no.2
• /
• pp.146-151
• /
• 2010

This paper proposes a rank-based control method of mutation probability for improving the performances of genetic algorithms (GAs). In order to improve the performances of GAs, GAs should not fall into premature convergence phenomena and should also be able to easily get out of the phenomena when GAs fall into the phenomena without destroying good individuals. For this, it is important to keep diversity of individuals and to keep good individuals. If a method for keeping diversity, however, is not elaborately devised, then good individuals are also destroyed. We should devise a method that keeps diversity of individuals and also keeps good individuals at the same time. To achieve these two objectives, we introduce a rank-based control method of mutation probability in this paper. We set high mutation probabilities to lowly ranked individuals not to fall into premature convergence phenomena by keeping diversity and low mutation probabilities to highly ranked individuals not to destroy good individuals. We experimented our method with typical four function optimization problems in order to measure the performances of our method. It was found from extensive experiments that the proposed rank-based control method could accelerate the GAs considerably.

• Journal of Digital Convergence
• /
• v.19 no.7
• /
• pp.271-282
• /
• 2021

This study was conducted to develop a model for predicting the length of stay for premature infants through machine learning. For the development of this model, 6,149 cases of premature infants discharged from the hospital from 2011 to 2016 of the discharge injury in-depth survey data collected by the Korea Centers for Disease Control and Prevention were used. The neural network model of the initial hospitalization was superior to other models with an explanatory power (R²) of 0.75. In the model added by converting the clinical diagnosis to CCS(Clinical class ification software), the explanatory power (R²) of the cubist model was 0.81, which was superior to the random forest, gradient boost, neural network, and penalty regression models. In this study, using national data, a model for predicting the length of stay for premature infants was presented through machine learning and its applicability was confirmed. However, due to the lack of clinical information and parental information, additional research is needed to improve future performance.

• Journal of Advanced Navigation Technology
• /
• v.12 no.5
• /
• pp.412-419
• /
• 2008

Genetic algorithms are efficient techniques for searching optimum solution but have the premature convergence problem getting stuck in the local optimum according to the evolutionary operator. In this paper we analyzed the reason for converging to the local optimum and proposed the method which able transit to the global optimum from the local optimum. In these methods we used the variable evolutionary operator with the average hamming distance, to maintain the genetic diversity of the population for getting out of the local optimum. The theoretical results are proved by the simulation experiments.

• 제어로봇시스템학회:학술대회논문집
• /
• 1993.10a
• /
• pp.853-858
• /
• 1993

The error back-propagation(BP) algorithm is widely used for finding optimum weights of multi-layer neural networks. However, the critical drawback of the BP algorithm is its slow convergence of error. The major reason for this slow convergence is the premature saturation which is a phenomenon that the error of a neural network stays almost constant for some period time during learning. An inappropriate selections of initial weights cause each neuron to be trapped in the premature saturation state, which brings in slow convergence speed of the multi-layer neural network. In this paper, to overcome the above problem, Micro-Genetic algorithms(.mu.-GAs) which can allow to find the near-optimal values, are used to select the proper weights and slopes of activation function of neurons. The effectiveness of the proposed algorithms will be demonstrated by some computer simulations of two d.o.f planar robot manipulator.