Shafaatunnur Hasan

Memetic binary particle swarm optimization for discrete optimization problems

In recent decades, many researchers have been interested in algorithms inspired by the observation of natural phenomena to solve optimization problems. Among them, meta-heuristic algorithms have been extensively applied in continuous (real) and discrete (binary) search spaces. Such algorithms are appropriate for global searches because of their global exploration and local exploitation abilities. In this study, a memetic binary particle swarm optimization (BPSO) scheme is introduced based on hybrid local and global searches in BPSO. The algorithm, binary hybrid topology particle swarm optimization (BHTPSO), is used to solve the optimization problems in the binary search spaces. In addition, a variant of the proposed algorithm, binary hybrid topology particle swarm optimization quadratic interpolation (BHTPSO-QI), is proposed to enhance the global searching capability. These algorithms are tested on two set of problems in the binary search space. Several nonlinear high-dimension functions and benchmarks for the 0-1 multidimensional knapsack problem (MKP) are employed to evaluate their performances. Their results are compared with some well-known modified binary PSO and binary gravitational search algorithm (BGSA). The experimental results showed that the proposed methods improve the performance of BPSO in terms of convergence speed and solution accuracy.

MPSO: Median-oriented Particle Swarm Optimization

Particle Swarm Optimization (PSO) is a bio-inspired optimization algorithm which has been empirically demonstrated to perform well on many optimization problems. However, it has two main weaknesses which have restricted the wider applications of PSO. The algorithm can easily get trapped in the local optima and has slow convergence speed. Therefore, improvement and/or elimination of these disadvantages are the most important objective in PSO research. In this paper, we propose Median-oriented Particle Swarm Optimization (MPSO) to carry out a global search over entire search space with accelerating convergence speed and avoiding local optima. The median position of particles and the worst and median fitness values of the swarm are incorporated in the standard PSO to achieve the mentioned goals. The proposed algorithm is evaluated on 20 unimodal, multimodal, rotated and shifted high-dimensional benchmark functions and the results are compared with some well-known PSO algorithms in the literature. The results show that MPSO substantially enhances the performance of the PSO paradigm in terms of convergence speed and finds global or good near-global optimal in the functions.

A differential-based harmony search algorithm for the optimization of continuous problems

We introduced a new harmony memory initialization method.We introduced a new pitch adjustment method based on DE/best/1 mutation strategy.We comprehensively studied the parameter setting of our algorithm.We compared our algorithm with the state of the art variants of HS algorithm.We compared our algorithm with the state of the art variants of DE algorithm. The performance of the Harmony Search (HS) algorithm is highly dependent on the parameter settings and the initialization of the Harmony Memory (HM). To address these issues, this paper presents a new variant of the HS algorithm, which is called the DH/best algorithm, for the optimization of globally continuous problems. The proposed DH/best algorithm introduces a new improvisation method that differs from the conventional HS in two respects. First, the random initialization of the HM is replaced with a new method that effectively initializes the harmonies and reduces randomness. Second, the conventional pitch adjustment method is replaced by a new pitch adjustment method that is inspired by a Differential Evolution (DE) mutation strategy known as DE/best/1. Two sets of experiments are performed to evaluate the proposed algorithm. In the first experiment, the DH/best algorithm is compared with other variants of HS based on 12 optimization functions. In the second experiment, the complete CEC2014 problem set is used to compare the performance of the DH/best algorithm with six well-known optimization algorithms from different families. The experimental results demonstrate the superiority of the proposed algorithm in convergence, precision, and robustness.

Multistrategy self-organizing map learning for classification problems

Multistrategy Learning of Self-Organizing Map (SOM) and Particle Swarm Optimization (PSO) is commonly implemented in clustering domain due to its capabilities in handling complex data characteristics. However, some of these multistrategy learning architectures have weaknesses such as slow convergence time always being trapped in the local minima. This paper proposes multistrategy learning of SOM lattice structure with Particle Swarm Optimisation which is called ESOMPSO for solving various classification problems. The enhancement of SOM lattice structure is implemented by introducing a new hexagon formulation for better mapping quality in data classification and labeling. The weights of the enhanced SOM are optimised using PSO to obtain better output quality. The proposed method has been tested on various standard datasets with substantial comparisons with existing SOM network and various distance measurement. The results show that our proposed method yields a promising result with better average accuracy and quantisation errors compared to the other methods as well as convincing significant test.

Soft Computing Methods for Big Data Problems

Generally, big data computing deals with massive and high-dimensional data such as DNA microarray data, financial data, medical imagery, satellite imagery, and hyperspectral imagery. Therefore, big data computing needs advanced technologies or methods to solve the issues of computational time to extract valuable information without information loss. In this context, generally, machine learning (ML) algorithms have been considered to learn and find useful and valuable information from large value of data. However, ML algorithms such as neural networks are computationally expensive, and typically, the central processing unit (CPU) is unable to cope with these requirements. Thus, we need a high-performance computer to execute faster solutions such graphics processing unit (GPU). GPUs provide remarkable performance gains compared to CPUs. The GPU is relatively inexpensive with affordable price, availability, and scalability. Since 2006, NVIDIA provides simplification of the GPU programming model with the Compute Unified Device Architecture (CUDA), which supports for accessible programming interfaces and industry-standard languages, such as C and C++. Since then, general-purpose graphics processing unit (GPGPU) using ML algorithms are applied on various applications, including signal and image pattern classification in biomedical area. The importance of fast analysis of detecting cancer or non-cancer becomes the motivation of this study. Accordingly, we proposed soft computing methods, self-organizing map (SOM) and multiple back-propagation (MBP) for big data, particularly on biomedical classification problems. Big data such as gene expression datasets are executed on high-performance computer and Fermi architecture graphics hardware. Based on the experiment, MBP and SOM with GPU-Tesla generate faster computing times than high-performance computer with feasible results in terms of speed and classification performance.

GPU-Based CAPSO with N-Dimension Particles

Today we are living in a world that is surrounded with information obesity which is also known as Big Data. Big data deals with zeta bytes of data flown from variety sources, and cannot be processed or analyzed using traditional procedure. Due to this, there is an increasing interest of researchers in using low cost GPUs for various applications that require intensive parallel computing to solve complex problems much faster. Various machine learning algorithms have been developed to obtain the optimal solutions with various data complexity. However, for big data problems, new machine learning algorithms need to be developed to deal with zeta bytes data problems. Centripetal accelerated particle swarm optimization (CAPSO) is the recent machine learning algorithm to enhance the convergence speed, accuracy and global optimality for optimization problems. However, the convergence speed of CAPSO is limited for small number of particles only. Hence, this research proposes improved CAPSO by implementing this algorithm on GPU platform through CUDA programming to handle N-dimensional scale of particles. Since CAPSO is intrinsically parallel processing, thus it can be effectively implemented on Graphics Processing Units (GPUs) according. The proposed GPU-based CAPSO was tested on various multi modal test functions and the results have proven that the proposed GPU-based CAPSO has successfully reduced the execution time with various particles dimensions compared to CPU-based CAPSO.

Faster Convergence of BP Network with Hybridization of Improved Cost Function and Control Memory Adaptation

Due to the weaknesses of Neural Network (NN) learning, this paper proposes an alternative approach in enhancing NN learning by integrating improved cost function with control adaptation of the nodes and address memory. As commonly known, weight adjustments of NN particularly in Back propagation (BP) algorithm, involve the connections between neurons, the activation function used by the neurons, the learning algorithm that specifies the procedure for adjusting the weights and the cost functions. The cost functions of BP are calculated based on the derivatives. These derivatives will determine the success rate of the application to train the network with an error function that resembles the objective of the problem on hand. Due to that, the concept of weights governance with control part mechanism between the input and hidden layer, and unit offsets of the hidden layer are implemented. to alleviate the problems of BP learning. The address memory part of the network will detain the output pattern of the hidden layer. Subsequently, the output patterns are compared with the input pattern, and propels back to the output layer after learning. From the experiments, we found that the results are promising with these mechanisms and improved cost function which yields faster convergence rates.