Fixed Length Chromosome Research Articles

Bridge the gap between fixed-length and variable-length evolutionary neural architecture search algorithms

<abstract><p>Evolutionary neural architecture search (ENAS) aims to automate the architecture design of deep neural networks (DNNs). In recent years, various ENAS algorithms have been proposed, and their effectiveness has been demonstrated. In practice, most ENAS methods based on genetic algorithms (GAs) use fixed-length encoding strategies because the generated chromosomes can be directly processed by the standard genetic operators (especially the crossover operator). However, the performance of existing ENAS methods with fixed-length encoding strategies can also be improved because the optimal depth is regarded as a known priori. Although variable-length encoding strategies may alleviate this issue, the standard genetic operators are replaced by the developed operators. In this paper, we proposed a framework to bridge this gap and to improve the performance of existing ENAS methods based on GAs. First, the fixed-length chromosomes were transformed into variable-length chromosomes with the encoding rules of the original ENAS methods. Second, an encoder was proposed to encode variable-length chromosomes into fixed-length representations that can be efficiently processed by standard genetic operators. Third, a decoder cotrained with the encoder was adopted to decode those processed high-dimensional representations which cannot directly describe architectures into original chromosomal forms. Overall, the performances of existing ENAS methods with fixed-length encoding strategies and variable-length encoding strategies have both improved by the proposed framework, and the effectiveness of the framework was justified through experimental results. Moreover, ablation experiments were performed and the results showed that the proposed framework does not negatively affect the original ENAS methods.</p></abstract>

A Variable-Length Chromosome Genetic Algorithm for Time-Based Sensor Network Schedule Optimization.

Scheduling sensor nodes has an important role in real monitoring applications using sensor networks, lowering the power consumption and maximizing the network lifetime, while maintaining the satisfaction to application requirements. Nevertheless, this problem is usually very complex and not easily resolved by analytical methods. In a different manner, genetic algorithms (GAs) are heuristic search strategies that help to find the exact or approximate global optimal solution efficiently with a stochastic approach. Genetic algorithms are advantageous for their robustness to discrete and noisy objective functions, as they are only evaluated at independent points without requirements of continuity or differentiability. However, as explained in this paper, a time-based sensor network schedule cannot be represented by a chromosome with fixed length that is required in traditional genetic algorithms. Therefore, an extended genetic algorithm is introduced with variable-length chromosome (VLC) along with mutation and crossover operations in order to address this problem. Simulation results show that, with help of carefully defined fitness functions, the proposed scheme is able to evolve the individuals in the population effectively and consistently from generation to generation towards optimal ones, and the obtained network schedules are better optimized in comparison with the result of algorithms employing a fixed-length chromosome.

A modified GA classifier for offline Tamil handwritten character recognition

In this paper, we proposed a classifier for Tamil handwritten character recognition using skeletonisation and modified GA to improve the recognition results of offline Tamil handwritten characters. The skeletonised character image is traversed from one endpoint to the other in an order, and based on the path of traversal, skeletonisation is explored to generate feature vector. The operations of conventional GA are modified to allow variable string length of chromosomes in GA. Fitness function is computed by integrating the classification capacity of the string metric Levenshtein distance which measures the dissimilarities between two strings. The experimental results on offline Tamil dataset demonstrates that the proposed classifier can automatically minimise the rate of misclassification and also provide better performance compared to GA with the fixed length chromosome. Our algorithm withstands even noisy data. Its comparison with other approaches is also substantiated and results proved that the proposed algorithm exhibits high accuracy in between 85% to 95%.

Reconfiguration of satellite orbit for cooperative observation using variable-size multi-objective differential evolution

A novel self-adaptive variable-size multi-objective differential evolution algorithm is presented to find the best reconfiguration of existing on-orbit satellites for some particular targets on the ground when an emergent requirement arises in a short period. The main contribution of this study is that three coverage metrics are designed to assess the performance of the reconfiguration. Proposed algorithm utilizes the idea of fixed-length chromosome encoding scheme combined with expression vector and the modified initialization, mutation, crossover and selection operators to search for optimal reconfiguration structure. Multi-subpopulation diversity initialization is adopted first, then the mutation based on estimation of distribution algorithm and adaptive crossover operators are defined to manipulate variable-length chromosomes, and finally a new selection mechanism is employed to generate well-distributed individuals for the next generation. The proposed algorithm is applied to three characteristically different case studies, with the objective to improve the performance with respect to specified targets by minimizing fuel consumption and maneuver time. The results show that the algorithm can effectively find the approximate Pareto solutions under different topological structures. A comparative analysis demonstrates that the proposed algorithm outperforms two other related multi-objective evolutionary optimization algorithms in terms of quality, convergence and diversity metrics.

Multiple parts process planning in serial–parallel flexible flow lines: part II—solution method based on genetic algorithms with fixed- and variable-length chromosomes

Multiple parts process planning (MPPP) is a hard optimization problem that requires the rigor and intensity of metaheuristic-based algorithms such as simulated annealing and genetic algorithms. In this paper, a solution method for this problem is developed based on genetic algorithms. Genetic algorithms solve problems by exploring a given search space. To do this, a landscape over which the search traverses is constructed based on a number of algorithm choices. Key algorithm choices include (a) type of chromosome representation, which affects the efficiency of an algorithm, and (b) type and form of genetic operators, which affect the effectiveness of an algorithm. More specifically, the suitability of a variable-length chromosome (VLC) representation for encoding a solution to a MPPP problem is investigated. The effectiveness and efficiency of implementing the VLC algorithm is analyzed and compared with: (a) the commonly used fixed-length chromosome representation, (b) a variant of the simulated annealing algorithm, and (c) a knowledge-informed simulated annealing algorithm. The scalability of the algorithms is analyzed and their effectiveness demonstrated by experimental results based on four problem sizes. Obtained results show that, although there are variances in performances, all algorithms investigated are capable of obtaining good solutions. In addition, variances were observed for different aspects of the MPPP problem. The results indicate that the VLC algorithm is effective in solving MPPP problems that consider multiple aspects in the search for optimal process planning solutions.

Automatic clustering method based on evolutionary optimisation

How to set the cluster number plays a key role in many clustering applications. To address this issue, this study introduces an automatic clustering method based on evolutionary algorithms (EAs). The basic idea is to convert a clustering problem into a global optimisation problem and tackle it by an EA. A new validity index, which balances the inter‐cluster consistency and the intra‐cluster consistency, is proposed to be the objective function. Three adaptive coding schemes, which can deal with variable‐length optimisation problems by using a fixed‐length chromosome, are designed to detect the cluster number automatically. The validity index and adaptive coding schemes are incorporated in an EA for automatic clustering. The authors approach is compared with some widely used validity indices and an adaptive coding scheme on some artificial data sets and two real‐world problems. The experimental results suggest that their method not only successfully detects the correct cluster numbers but also achieve stable results for most of test problems.

Road traffic control based on genetic algorithm for reducing traffic congestion

AbstractIn this paper, we propose a road traffic control method for reducing traffic congestion with a genetic algorithm. In the not too distant future, a system which controls the routes of all vehicles in a certain area must be realized. The system should optimize the routes of all vehicles, but the solution space of this problem is enormous. Therefore, we apply the genetic algorithm to this problem by encoding the route of all vehicles to a fixed length chromosome. To improve the search performance, a new genetic operator called “path shortening” is also designed. The effectiveness of the proposed method is shown experimentally. © 2012 Wiley Periodicals, Inc. Electron Comm Jpn, 95(4): 11‐19, 2012; Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/ecj.10421

Efficient Constellation Design Based on Improved Non-dominated Sorting Genetic Algorithm-II

Constellation design is a typical multiple peaks, multiple valleys and non-linear multi-objective optimization problem. How to design satellite constellation is one of the key sectors of research in the aerospace field. In this paper, in order to improve the global convergence and diversity performance of traditional constellation optimization algorithm, multi-parent arithmetic crossover and SBX crossover operator of NSGA-II are used to improve searching capability of this algorithm. Meanwhile, Gaussian mutation and Cauchy mutation, with diversity of population, make the algorithm get better behaviors in convergence and diversity of finding solutions. Based on the methods, an improvement NSGA-II is presented to design constellation in the paper. The algorithm uses fixed length chromosome representation. Real coding is adopted for that the problem has both integer continuous variables. Combining the coverage assessment criterions, an orbit parameters optimization framework based on non-dominated sorting genetic algorithm (NSGA-II) was proposed. This method is applied to a detailed example, and result shows that a group of Pareto solutions with good spread can be achieved, which gives strong support to constellation scheme determination.

A Novel Implementation of G-Fuzzy Logic Controller Algorithm on Mobile Robot Motion Planning Problem

This paper presents a new algorithm for global path planning to a goal for a mobile robot using GA and fuzzy Algorithms. A genetic algorithm is used to find the optimal path for a mobile robot to move in a dynamic environment expressed by a map with nodes and links. Locations of target and obstacles to find an optimal path are given in an environment that is a 2-D workplace. Each via point (landmark) in the net is a gene which is represented using binary code. The number of genes in one chromosome is function of the number of obstacles in the map. Therefore, we used a fixed length chromosome. The generated robot path is optimal in the sense of the shortest distance. The fitness function of genetic algorithm takes full consideration of three factors: the collision avoidance path, the shortest distance and smoothness of the path. The specific genetic operators are also selected to make the genetic algorithm more effective. The simulation results verify that the genetic algorithm is high effective under various complex dynamic alien environments.

交通渋滞緩和のための遺伝的アルゴリズムに基づく道路交通管制

In this paper, we propose a road traffic control method for reducing traffic congestion with a genetic algorithm. In the not too distant future, a system which controls the routes of all vehicles in a certain area must be realized. The system should optimize the routes of all vehicles, but the solution space of this problem is enormous. Therefore, we apply the genetic algorithm to this problem by encoding the route of all vehicles to a fixed length chromosome. To improve the search performance, a new genetic operator called “path shortening” is also designed. The effectiveness of the proposed method is shown experimentally. © 2012 Wiley Periodicals, Inc. Electron Comm Jpn, 95(4): 11-19, 2012; Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/ecj.10421

Evolutionary Path Planning With Subpath Constraints

We develop an evolutionary method of planning paths that are subject to subpath constraints. These constraints can include subpaths that must be incorporated into the solution path, path intersection restrictions, and obstacle avoidance. Our method involves two stages. In the first stage, a global solution is found without consideration of any obstacles. In the second stage, local planning is performed to modify the global path to avoid obstacles. Stage one involves a fixed-length chromosome formulation of a genetic algorithm that utilizes existing operators and a new subpath reversal operator. Stage two involves a graph search process. Our algorithm is applicable to the field of printed electronics where continuous-spray processes are used to deposit electrically functional material onto flexible substrates. These processes give rise to the kinds of subpath constraints we have investigated. We evaluate our algorithm by applying it to a representative problem in the printed electronics field.

An integer-coded evolutionary approach for mixture maximum likelihood clustering

This paper outlines an algorithm for solving the maximum mixture likelihood clustering problem using an integer-coded genetic algorithm (IGA-ML) where a fixed length chromosome encodes the object-to-cluster assignment. The main advantage of the outlined algorithm (IGA-ML) compared with other known algorithms, such as the k-means technique, is that it can successfully discover the correct number of clusters, in addition to carrying out the partitioning process. The algorithm implements a post-fixing sorting mechanism that drastically reduces the searched solution space by eliminating duplicate solutions that appear after applying the genetic operations. Simulation results show the effectiveness of the algorithm especially with the case of overlapping clusters.

Optimal Design of a Polarization-Transformation Filter Using GAs with Variable- and Fixed-Length Chromosomes

Optimal Design of a Polarization-Transformation Filter Using GAs with Variable- and Fixed-Length Chromosomes