Papers by Orawan WATCHANUPAPORN
ABSTRACT This paper proposes a new evolutionary algorithm called LZWMIMIC. The proposed algorithm... more ABSTRACT This paper proposes a new evolutionary algorithm called LZWMIMIC. The proposed algorithm combines the LZW compressed chromosome encoding and Mutual-Information-Maximizing Input Clustering (MIMIC) algorithm. The advantage of LZW encoding is that it reduces the search space thus speeds up the evolutionary search. The advantage of MIMIC is that it can solve complex problem by finding a relationship between gene positions. The performance of the original MIMIC and LZWMIMIC are compared on standard benchmark problems. Further, compressed chromosome length and problem size are varied to see their effect in the performance. The experimental results show that our proposed algorithm outperforms the original MIMIC.
LZWGA is an algorithm that combines LZW compression algorithm with genetic algorithm (GA). An LZW... more LZWGA is an algorithm that combines LZW compression algorithm with genetic algorithm (GA). An LZWGA chromosome can be decompressed to a GA binary-string chromosome. In this paper, we propose a new mutation operator for LZWGA called conditional probability mutation (CPM). In contrast to original LZWGA mutation which randomly changes some values in an individual, CPM takes advantage of the relationship between gene positions. We compare the performance of LZWGA with original mutation and LZWGA with CPM on non random and random version of standard benchmark problems. Furthermore, we vary mutation rate to see its effect in the performance. The experimental results show that our proposed mutation outperforms original LZWGA mutation in non random problems.
In this paper, a group of mobile robots learns to solve a target reaching problem in a simulated ... more In this paper, a group of mobile robots learns to solve a target reaching problem in a simulated grid environment filled with obstacles. Each robot knows its distance to the target and can communicate with each other. The proposed learning algorithm combines a reinforcement learning algorithm and a swarm optimization algorithm. Q-learning, which is a reinforcement learning algorithm, is modified to learn a policy by specifying rewards and punishment for certain robot actions. Particle swarm optimization (PSO), which is a swarm optimization algorithm, is modified for grid environment and used to accelerate the learning process for multiple robots. The proposed algorithm outperforms the original Q-learning in both training and testing. It learns 2.23 times faster and required 6.52 fewer steps to reach the destination. Moreover, it uses less memory than the original Q-learning. We also experiment on various numbers of robots. The result shows that more robots learn faster in most cases.
Journal of Convergence Information Technology, Jul 31, 2013
Advanced Science Letters, 2014
International Journal of Applied Evolutionary Computation, Oct 1, 2013
Abstract-This paper proposes a new evolutionary algorithm called LZWCGA. LZWCGA is an algorithm t... more Abstract-This paper proposes a new evolutionary algorithm called LZWCGA. LZWCGA is an algorithm that combines the LZW compressed chromosome encoding and compact genetic algorithm (cGA). The advantage of LZW encoding is to reduce the search space thus speed up the evolutionary search. cGA is one of Estimation of Distribution Algorithms. Its advantage is compact representation of the whole binary-string genetic algorithm population.
GSTF Journal on computing, 2014
Differential Evolution (DE) is a fast and robust real vector optimizer. This paper applies DE to ... more Differential Evolution (DE) is a fast and robust real vector optimizer. This paper applies DE to discrete problems by converting a real chromosome to an integer chromosome and then decompress to a binary chromosome using LZW algorithm. Experimental result shows that this approach is better than the previous work and the evolution time is very fast. Analysis result shows that the fitness landscape of LZW encoding is less complex than the original encoding for each test problem.
2016 2nd International Conference on Control, Automation and Robotics (ICCAR), 2016
In this paper, a group of mobile robots learns to solve a target reaching problem in a simulated ... more In this paper, a group of mobile robots learns to solve a target reaching problem in a simulated grid environment filled with obstacles. Each robot knows its distance to the target and can communicate with each other. The proposed learning algorithm combines a reinforcement learning algorithm and a swarm optimization algorithm. Q-learning, which is a reinforcement learning algorithm, is modified to learn a policy by specifying rewards and punishment for certain robot actions. Particle swarm optimization (PSO), which is a swarm optimization algorithm, is modified for grid environment and used to accelerate the learning process for multiple robots. The proposed algorithm outperforms the original Q-learning in both training and testing. It learns 2.23 times faster and required 6.52 fewer steps to reach the destination. Moreover, it uses less memory than the original Q-learning. We also experiment on various numbers of robots. The result shows that more robots learn faster in most cases.
2012 Sixth International Conference on Genetic and Evolutionary Computing, 2012
2011 4th International Conference on Biomedical Engineering and Informatics (BMEI), 2011
ABSTRACT This paper proposes a new evolutionary algorithm called LZWMIMIC. The proposed algorithm... more ABSTRACT This paper proposes a new evolutionary algorithm called LZWMIMIC. The proposed algorithm combines the LZW compressed chromosome encoding and Mutual-Information-Maximizing Input Clustering (MIMIC) algorithm. The advantage of LZW encoding is that it reduces the search space thus speeds up the evolutionary search. The advantage of MIMIC is that it can solve complex problem by finding a relationship between gene positions. The performance of the original MIMIC and LZWMIMIC are compared on standard benchmark problems. Further, compressed chromosome length and problem size are varied to see their effect in the performance. The experimental results show that our proposed algorithm outperforms the original MIMIC.
International Journal of Applied Evolutionary Computation, 2013
Estimation of distribution algorithm (EDA) can solve more complicated problems than its predecess... more Estimation of distribution algorithm (EDA) can solve more complicated problems than its predecessor (Genetic Algorithm). EDA uses various methods to probabilistically model a group of highly fit individuals. Calculating the model in sophisticated EDA is very time consuming. To reduce the model building time, the authors propose compressed chromosome encoding. A chromosome is encoded using a format that can be decompressed by the Lempel-Ziv-Welch (LZW) algorithm. The authors combined LZW encoding with various EDAs and termed the class of algorithms Lempel-Ziv-Welch Estimation of Distribution Algorithms (LZWEDA). Experimental results show that LZWEDA significantly outperforms the original EDA. Finally, the authors analyze how LZW encoding transforms a fitness landscape.
Journal of Convergence Information Technology, 2013
Proceedings of the International Conference on Management of Emergent Digital EcoSystems, 2010
LZWGA is an algorithm that combines LZW compression algorithm with genetic algorithm (GA). An LZW... more LZWGA is an algorithm that combines LZW compression algorithm with genetic algorithm (GA). An LZWGA chromosome can be decompressed to a GA binary-string chromosome. In this paper, we propose a new mutation operator for LZWGA called conditional probability mutation (CPM). In contrast to original LZWGA mutation which randomly changes some values in an individual, CPM takes advantage of the relationship between gene positions. We compare the performance of LZWGA with original mutation and LZWGA with CPM on non random and random version of standard benchmark problems. Furthermore, we vary mutation rate to see its effect in the performance. The experimental results show that our proposed mutation outperforms original LZWGA mutation in non random problems.
Advanced Science Letters, 2014
ECTI Transactions on Computer and Information Technology (ECTI-CIT), 1970
Compressed compact genetic algorithm (c2GA) is an algorithm that utilizes the compressed chromoso... more Compressed compact genetic algorithm (c2GA) is an algorithm that utilizes the compressed chromosome encoding and compact genetic algorithm (cGA). The advantage of c2GA is to reduce the memory usage by representing population as a probability vector. In this paper, we analyze the performance in term of robustness of c2GA. Since the compression and decompression strategy employ two parameters, which are the length of repeating value and the repeat count, we vary these two parameters to see the performance affected in term of convergence speed. The experimental results show that c2GA outperforms cGA and is a robust algorithm.
Uploads
Papers by Orawan WATCHANUPAPORN