Inadequately labeled data can limit the accuracy of classification in image recognition tasks. Several methods have been proposed in the past to alleviate this limitation, such as transfer learning and data augmentati...
详细信息
ISBN:
(纸本)9781450362948
Inadequately labeled data can limit the accuracy of classification in image recognition tasks. Several methods have been proposed in the past to alleviate this limitation, such as transfer learning and data augmentation. However, the classification accuracy of the convolutional neural network (CNN) largely depends on its structural parameters, which are known as hyper-parameters. Therefore, in this paper, we introduce another method for minimizing the misclassification rate in a given small dataset by determining the hyper-parameters. The harmony search (HS) algorithm, improved harmony search (IHS) algorithm, self-adaptive global best harmony search (SGHS) algorithm, and novel global harmony search (NGHS) algorithm are applied for determining the optimal hyper-parameters. Additionally, we also compared the estimation performances of these four HS algorithms. It was finally observed that the HS and the IHS algorithms greatly outperform the other two algorithms.
Digital IIR filter design by optimizing a fitness function with respect to coefficients of a filter with rational transfer function by meta-heuristic algorithms has been considered recently, Most researchers use a fit...
详细信息
ISBN:
(纸本)9781509043309
Digital IIR filter design by optimizing a fitness function with respect to coefficients of a filter with rational transfer function by meta-heuristic algorithms has been considered recently, Most researchers use a fitness function consisted of difference between magnitude response of desired filter and designed filter and the constraints such as linear phase, minimum phase and stability of designed filter. In this paper, a comprehensive fitness function for IIR digital filter design with 6 terms is proposed. A new term is added to fitness function to get a filter with low delay. Low delay filters are desirable for real time signal processing. This term is weighted partial energy of the impulse response of designed causal filter. Maximizing this term leads to concentration of energy of impulse response at its beginning, consequently a low delay filter. Low delay property leads to fast decaying of transient response and low delay between input and output of designed filter. Proposed fitness function also includes some terms to meet linear phase, minimum phase and stability constraints. Meta-heuristic optimization algorithms GA, GSA and PSO are used to optimize proposed fitness function. To evaluate efficiency of the proposed method, it will be used to design a low delay low pass filter and a low delay differentiator. Reported results show lower delay of designed filters by proposed method than designed ones by traditional methods.
This paper aims to discuss and compare various metaheuristic algorithms applied to the "Knapsack Problem". The Knapsack Problem is a combinatorial optimization maximization problem which requires to find the...
详细信息
ISBN:
(纸本)9781509035199
This paper aims to discuss and compare various metaheuristic algorithms applied to the "Knapsack Problem". The Knapsack Problem is a combinatorial optimization maximization problem which requires to find the number of each weighted item to be included in a hypothetical knapsack, so the total weight is less than or equal to the required weight. To come to an optimized solution for such a problem, a variety of algorithms can possibly be used. In this paper, Tabu Search, Scatter Search and Local Search algorithms are compared taking execution time, solution quality and relative difference to best known quality, as metrics to compute the results of this NP-hard problem.
Generally speaking, it is not fully understood why and how metaheuristic algorithms work very well under what conditions. It is the intention of this paper to clarify the performance characteristics of some of popular...
详细信息
ISBN:
(纸本)9783662479261;9783662479254
Generally speaking, it is not fully understood why and how metaheuristic algorithms work very well under what conditions. It is the intention of this paper to clarify the performance characteristics of some of popular algorithms depending on the fitness landscape of specific problems. This study shows the performance of each considered algorithm on the fitness landscapes with different problem characteristics. The conclusions made in this study can be served as guidance on selecting algorithms to the problem of interest.
In this paper, the linear quadratic regulator (LQR) optimal control of constrained dynamic systems for generalized systems with external disturbances is designed. At first, a generalized system is given, this model in...
详细信息
ISBN:
(纸本)9781728172965
In this paper, the linear quadratic regulator (LQR) optimal control of constrained dynamic systems for generalized systems with external disturbances is designed. At first, a generalized system is given, this model includes some ordinary differential equations (ODEs) and some algebraic equations as a constraint on dynamics. In this paper, we address the optimal LQR control for continuous linear generalized systems in which the system has only finite dynamic modes. In the LQR problem, we derive a Riccati equation in each case. Then a linear disturbance observer is proposed to estimate the disturbance. Then, we propose four types of metaheuristic algorithms such as grasshopper optimization algorithm (GOA), grey wolf optimizer (GWO), genetic algorithm (GA), and particle swarm optimization (PSO) with a suitable structure for solving Riccati equation. In each case, the solution obtained by GOA, GWO, GA, and PSO. Simulation results demonstrate the good performance of the proposed methods.
Buildings consume approximately 40% of end-use energy worldwide and are responsible for approximately one-third of greenhouse gas (GHG) emissions. Clearly, designing high energy performance buildings and identifying e...
详细信息
Buildings consume approximately 40% of end-use energy worldwide and are responsible for approximately one-third of greenhouse gas (GHG) emissions. Clearly, designing high energy performance buildings and identifying effective energy retrofit measures not only decrease CO 2 emissions, but also reduce the need for non-renewable energy sources. While the traditional rules of thumb and building codes improve the building energy efficiency, they are likely to be far from the optimal design as they do not consider the interactions among design variables. Therefore, new methods should be developed to achieve the maximum energy savings. Building energy optimisation (BEO) is a method that considers interactions among design variables and selects the optimal building design from a set of available alternative designs based on the mathematics. A challenge of currently-available optimisation methods is that they suffer from high computational cost due to high complexities in building optimisation problems including multi-modal and nonlinear behaviour of building thermal performance, discontinuities in the optimisation variables (e. g. window type), uncertainty in building design parameters (e. g. alterations in building operating conditions) and discontinuities in the output of building simulation software (e. g. EnergyPlus). This high computational cost remains a key barrier to the widespread utilisation of optimisation as a design tool. Accordingly, the focus of this research is on developing new efficient solution methods for Building Optimisation Problems (BOPs) and deploying them on realistic case studies to evaluate their performance and utility. Generally, BOPs can be categorised into two main groups: simulation-based optimisation (software-in-the- loop method) and surrogate-based optimisation methods. In this thesis, new methods were developed to improve the performance of both methods. Furthermore, a new methodology was developed to address uncertainty of building simu
Single-nucleotide polymorphism (SNP) analysis has become a pivotal strategy for drug discovery within bioinformatics, especially for incurable diseases like cancer. With the increasing number of researchers starting t...
详细信息
In recent years, Wireless Sensor Network (WSN) is in demand over the world due to its rapid deployment in a variety of applications. The challenges in the WSN like limited battery power and communication range caught ...
详细信息
This paper presents a comparative study of five metaheuristic algorithms, namely, salp swarm algorithm (SSA), owl search algorithm (OSA), sooty tern optimization algorithm (STOA), squirrel search algorithm (SqSA), and...
详细信息
Timetabling is something that a lot of businesses, public sectors and institutions have to deal with, which is probably why automating the process of creating good timetables is a well-known problem that has been well...
详细信息
Timetabling is something that a lot of businesses, public sectors and institutions have to deal with, which is probably why automating the process of creating good timetables is a well-known problem that has been well researched during the re- cent decades. This thesis introduces a timetabling problem named project matching problem, which is a problem of assigning workers to projects and schedule these projects into a limited time frame, while respecting certain constraints such as the workers preference towards these projects. In order to efficiently solve instances of this problem, an application framework based on the metaheuristic algorithm Iter- ated local search is developed and implemented using common software techniques and architectural patterns. As the project matching problem has a broad definition, the idea is that the area of use for this application framework comprises timetabling problems outside the domain of project organisation. The results are based on running these problem instances through the application framework, where the results of these runs are presented in terms of both solution quality and algorithmic efficiency. The solution quality is based on analyses of the actual solutions that is produced while algorithmic efficiency is measured on the execution time of the algorithm, which offers a suggestion of how the solution quality improves as the execution time increases. For one of the problem instances this application framework produced an optimal solution, which offers an indication on the applicability of this framework.
暂无评论