Comparative analysis of optimization algorithm-based training algorithms for a few shot continual learning

PurposeSea Lion Optimization (SLnO) algorithm involves the ability of exploration and exploitation phases, and it is able to solve combinatorial optimization problems. For these reasons, it is considered a global optimizer. The scheduling operation is completed by imitating the hunting behavior of sea lions.Design/methodology/approachCloud computing (CC) is a type of distributed computing, contributory in a massive number of available resources and demands, and its goal is sharing the resources as services over the internet. Because of the optimal using of these services is everlasting challenge, the issue of task scheduling in CC is significant. In this paper, a task scheduling technique for CC based on SLnO and multiple-objective model are proposed. It enables decreasing in overall completion time, cost and power consumption; and maximizes the resources utilization. The simulation results on the tested data illustrated that the SLnO scheduler performed better performance than other state-of-the-art schedulers in terms of makespan, cost, energy consumption, resources utilization and degree of imbalance.FindingsThe performance of the SLnO, Vocalization of Whale Optimization Algorithm (VWOA), Whale Optimization Algorithm (WOA), Grey Wolf Optimization (GWO) and Round Robin (RR) algorithms for 100, 200, 300, 400 and 500 independent cloud tasks on 8, 16 and 32 VMs was evaluated. The results show that SLnO algorithm has better performance than VWOA, WOA, GWO and RR in terms of makespan and imbalance degree. In addition, SLnO exhausts less power than VWOA, WOA, GWO and RR. More precisely, SLnO conserves 5.6, 21.96, 22.7 and 73.98% energy compared to VWOA, WOA, GWO and RR mechanisms, respectively. On the other hand, SLnO algorithm shows better performance than the VWOA and other algorithms. The SLnO algorithm's overall execution cost of scheduling the cloud tasks is minimized by 20.62, 39.9, 42.44 and 46.9% compared with VWOA, WOA, GWO and RR algorithms, respectively. Finally, the SLnO algorithm's average resource utilization is increased by 6, 10, 11.8 and 31.8% compared with those of VWOA, WOA, GWO and RR mechanisms, respectively.Originality/valueTo the best of the authors’ knowledge, this work is original and has not been published elsewhere, nor is it currently under consideration for publication elsewhere.

Centroid mutation-based Search and Rescue optimization algorithm for feature selection and classification

Optimization is a vital constituent of machine learning as it brings about efficiency of the analysis process. An optimizer is used to minimize an objective function, which in turn leverages efficiency of any given model. Many metaheuristic algorithms have been applied on many models for optimization purposes. The mostly favoured metaheuristic algorithms are the population-based ones as they initialize a population of candidate solutions, rather than a single-solution strategy. This paper takes off from the original research made on fuzzy C-means (FCM), where the weighting component (m) was set at 2. Setting the weighting component at 2 may not ideally work well for all applications. Then, the point of this examination is to optimize the weighting component m using bio-inspired algorithms that are population-based in nature. The idea is to compare performance of conventional FCM with improved version of FCM by augmenting it with different optimization techniques. Three comparatively new and popular bio-inspired optimization algorithms such as moth-flame Optimizer (MFO), gray wolf optimizer (GWO) and whale optimization algorithm (WOA) are selected to optimize tunable parameters of conventional FCM. For experimentation purpose, five different data sets namely seeds, iris, glass identification, mall customer segmentation and BuddyMove are selected. The experimental results show considerable improvements in the overall performance of conventional FCM as a consequence of parameter optimization. A new approach for optimizing FCM parameters using bio-inspired algorithms is presented.

Open-source software is gaining popularity in industrial projects due to its accessibility and cost-effectiveness. However, concerns persist about its quality and reliability. To assess software reliability quantitatively, software reliability models are utilized, with the unknown parameters of these models typically determined using statistical techniques. In many cases, these methods fail to converge to the global optimal solution of parameter estimation of nonlinear mathematical models and are quite sensitive to the initial guesses of unknown parameters. This necessitates employing a high-quality parameter estimation technique. The study demonstrates the potential application of nine nature-inspired swarm intelligence-based algorithms to address nonlinear parameter estimation problems and effectively identify the global optimal solution with high likelihood, irrespective of the initial guess. These typical algorithms are classified into several categories, including animal-inspired algorithms such as grey wolf optimizer, insect-inspired algorithms such as artificial bee colony, social spider optimization, firefly algorithm, and moth flame optimization, bird-inspired algorithms such as particle swarm optimization, sea creature-inspired algorithms such as whale optimization algorithm, and plant-inspired algorithms such as flower pollination algorithm and dandelion optimizer. Three real-world, open-source reliability datasets are utilized to assess the efficacy of these algorithms in estimating the parameters of two prominent non-homogeneous Poisson process models in software reliability.

In evaluating solar photovoltaic (SPV) cell performance and monitoring operational deviations, parameters of the solar cell equivalent circuit models play an important role. Given that solar cells have nonlinear current-voltage characteristics, calculating their parameters is a significant challenge. Therefore, to effectively handle this engineering challenge, an accurate and efficient optimization technique is typically needed. In order to determine SPV cell parameters, this study revealed a new optimization technique called Ali Baba and the Forty Thieves (AFT) algorithm. The suggested optimization technique was used to estimate the parameters of the single-diode equivalent circuit model of the SPV. Study work has been done on various photovoltaic modules, including the Photowatt PWP-201 and the RTC France solar cell. In addition, a comparison study is used to demonstrate the proposed AFT algorithm's performance against a number of existing heuristic algorithms, including the Moth Flame Optimization (MFO), Dragonfly Algorithm (DA), Whale Optimization Algorithm (WOA), Grey Wolf Optimization (GWO), Ant Lion Optimization (ALO), Harris Hawk Optimization (HHO), Hybrid of Particle Swarm Optimization and Grey Wolf Optimization (PSOGWO), Marine Predator Algorithm (MPA), and African Vulture Optimization Algorithm (AVOA). The comparison has been made using the same datasets and the same computational workload for each optimization technique in order to evaluate performance fairly. The acquired comparative results demonstrated that the AFT algorithm provides higher performance than any of those techniques in terms of the root mean square error (RMSE), computation time, and the accuracy of the results for parameter estimation of SPV cells.

The global demand for electrical energy has witnessed a substantial increase, presenting a challenge for power systems worldwide. In addition to technical considerations, the escalating issue of global warming has become a paramount concern in the planning studies of various sectors. The formulation and resolution of a single-objective non-linear optimization problem are carried out, considering different operational scenarios. Recent heuristic algorithms, including Particle Swarm Optimization (PSO), Cat Swarm Optimization (CSO), Teaching Learning Based Optimization (TLBO), Grey Wolf Optimization (GWO) and Chimp Optimization algorithm (ChOA) are employed to address the complexities associated with maximizing power output under partial shading conditions in solar PV systems. The inherent challenges of achieving MPPT under such conditions make conventional analytic approaches computationally intensive. Hence, this study leverages heuristic algorithms to optimize solar PV system performance, providing efficient solutions to the associated optimization problems. The current research work was performed on a test system using a MATLAB/SIMULINK environment and the results are presented and discussed. From the simulation results, it was found that ChOA have shown higher conversion efficiency of 99.63% with maximum power output of 525.13 W when compared to other optimization algorithms for the given shading pattern condition. Further, ChOA offers easy implementation and faster convergence, outperforming established methods in GMPP search by reducing power oscillations and achieving precise MPP convergence.

In this study, the performances of eight recent swarm intelligence techniques, accelerated particle swarm optimization (APSO), firefly algorithm, levy-flight krill herd, whale optimization algorithm (WOA), ant lion optimizer, grey wolf optimizer, moth-flame optimization algorithm and teaching-learning-based optimization algorithm (TLBO), are explored. Particle swarm optimization algorithm is also considered to benchmark the efficiencies. A final cost is considered as an objective function which deals with shallow footing optimization with two attitudes: routine optimization, and sensitivity analysis. Moreover, as a further study, the effect of the location of the column at the top of the foundation is examined by adding two spare design variables. To this end, three numerical case studies are simulated. Based on the final results TLBO showed an acceptable performance because of the lowest mean values and WOA demonstrated the weakest efficiency among the algorithms in this study.

One of the most parameters and variables in every economics is the interest rate. Government officials and lawmakers change interest rates for various purposes: controlling liquidity, inflation, and prices, Economic growth and development, lending, etc. So, it is important to set the interest rate correctly. If you can predict the interest rate correctly, you can earn and gain profit by investing in various sectors. Moreover, the interest rate can impact other sectors through parallel markets such as the stock market, automobile, housing, etc. Interest rates are related to parallel markets. Thus, if you can forecast the interest rate, you can predict the parallel markets too. The main goal of this article, as it is clear from the title, is the prediction of interest rate using ANN and improving the network using some novel heuristic algorithms such as Moth Flame Optimization algorithm (MFO), Chimp Optimization Algorithm (CHOA), Time-varying Correlation Particle Swarm Optimization algorithm (TVAC-PSO), etc. we used 17 variables such as oil price, gold coin price, house price, etc. as input variables. We used GA and a new algorithm called Grey Wolf Optimization, Particle Swarm Optimization (GWO-PSO) algorithm as a feature selection and choosing the best variables. We have used eight loss functions such as MSE, RMSE, MAE, etc. too. Finally, we have compared different algorithms due to their estimation errors. The main contribution of this paper is that, first, this is for the first time which these novel metaheuristic algorithms have been used for the prediction of interest rate. Second, it has tried to use different graphs and tables for better understanding and totally a comprehensive research paper. The results show that Whale Optimization Algorithm (WOA) performed better than other methods along with less error.

This article presents a novel hybrid algorithm based on the grey-wolf optimizer and whale optimization algorithm, referred here as grey-wolf optimizer–whale optimization algorithm, for the dynamic parametric modelling of a four degree-of-freedom master–slave robot manipulator system. The first part of this work consists of testing the feasibility of the grey-wolf optimizer–whale optimization algorithm by comparing its performance with a grey-wolf optimizer, whale optimization algorithm and particle swarm optimization using 10 benchmark functions. The grey-wolf optimizer–whale optimization algorithm is then used for the model identification of an experimental master–slave robot manipulator system using the autoregressive moving average with exogenous inputs model structure. Obtained results demonstrate that the hybrid algorithm is effective and can be a suitable substitute to solve the parameter identification problem of robot models.

Classification and rapid non-destructive quality evaluation of different processed products of Cyperus rotundus based on near-infrared spectroscopy combined with deep learning

Optimization algorithms are one of the effective stochastic methods in solving optimization problems. In this paper, a new swarm-based algorithm called Northern Goshawk Optimization (NGO) algorithm is presented that simulates the behavior of northern goshawk during prey hunting. This hunting strategy includes two phases of prey identification and the tail and chase process. The various steps of the proposed NGO algorithm are described and then its mathematical modeling is presented for use in solving optimization problems. The ability of NGO to solve optimization problems is evaluated on sixty-eight different objective functions. To analyze the quality of the results, the proposed NGO algorithm is compared with eight well-known algorithms, particle swarm optimization, genetic algorithm, teaching-learning based optimization, gravitational search algorithm, grey wolf optimizer, whale optimization algorithm, tunicate swarm algorithm, and marine predators algorithm. In addition, for further analysis, the proposed algorithm is also employed to solve four engineering design problems. The results of simulations and experiments show that the proposed NGO algorithm, by creating a proper balance between exploration and exploitation, has an effective performance in solving optimization problems and is much more competitive than similar algorithms.

This paper introduces an application of the dandelion optimization (DO) algorithm in antenna arrays. This is the first time that the DO algorithm has been used for optimizing antenna arrays. For antenna array optimization, sidelobe level (SLL) and deep nulls are key technical indicators. A lower SLL can improve the signal-to-noise ratio and reduce the impact of clutter signals outside the main beam. Deep nulls need to be aligned with the direction of interference to eliminate the influence of interference sources. The combination of the two can effectively improve the anti-interference ability of the entire system. Therefore, antenna arrays with ultra-low sidelobes and ultra-deep nulls are currently hot in the field of antenna array design and are also some of the key technologies needed to achieve modern high-performance radar systems. As a new type of evolutionary algorithm inspired by nature, the DO algorithm is inspired by the wind propagation behavior of dandelions in nature. This algorithm iteratively updates the population from three stages of ascent, descent, and landing, ultimately finding the optimal position. It has good optimization ability in solving complex problems such as those involving nonlinearity, discreteness, and non-convexity, and the antenna array pattern synthesis optimization problem belongs to multivariate nonlinear problems. Therefore, the DO algorithm can be effectively applied in the field of antenna array optimization. In this work, we use the following method to obtain an optimized pattern of a linear array with the lowest sidelobe level (SLL), null placement in particular directions, and a lower notch in particular directions: by controlling the antenna array’s element spacing and leaving the phase unchanged to optimize the current amplitudes and by controlling the excitation current and phase fixation of the antenna array and changing the element spacing. In the first and second examples, different algorithms are used to reduce the SLL of the antenna. In the first example, the DO algorithm reduces the SLL to −33.37 dB, which is 2.67 dB, 2.67 dB, 3.77 dB, 2.74 dB, and 2.52 dB lower than five other algorithms. In the second example, the SLL optimized by the DO algorithm is −42.56 dB, which is 5.04 dB and 1.48 dB lower than two other algorithms. In both examples, the DO algorithm reduces the SLL lower than other algorithms when the main lobe of the antenna is not significantly widened. Examples 3, 4, and 5 use the DO algorithm to optimize the amplitude of the current, generating deep nulls and deep notches in specific directions. In Example 3, the DO algorithm obtains a depth of nulls equal to −187.6 dB, which is 66.7 dB and 44.3 dB lower than that of the flower pollination algorithm (FPA) and the chaotic colony predation algorithm (CCPA), respectively. In Example 4, the deep null obtained by the DO algorithm is as low as −98.69 dB, which is 6.67 dB lower than the deep null obtained by the grey wolf optimization (GWO) algorithm. In Example 5, the deep notch obtained by the DO algorithm is as low as −63.1 dB, which is 6.4 dB and 1.9 dB lower than the spider monkey optimization (SMO) algorithm and the grasshopper optimization algorithm (GOA), respectively. The data prove that the DO algorithm produces deeper nulls and notches than other algorithms. The last two examples involve reducing sidelobe levels and generating deep nulls by optimizing the spacing between elements. In Example 5, the SLL obtained using the DO algorithm is −22.8766 dB, which is 0.1998 dB lower than the lowest SLL of −22.6768 dB among other algorithms. In Example 6, the SLL obtained using the DO algorithm is −20.1012 dB, and the null depth is −125.1 dB, which is 1.592 dB lower than the SLL obtained by the cat swarm optimization (CSO) algorithm and 19.1 dB lower than the deep null obtained by the GWO algorithm, respectively. In summary, the results of six simulation experiments indicate that the DO algorithm has better optimization ability in linear array optimization than other evolutionary algorithms.

In the field of asset allocation, how to balance the returns of an investment portfolio and its fluctuations is the core issue. Capital asset pricing model, arbitrage pricing theory, and Fama–French three-factor model were used to quantify the price of individual stocks and portfolios. Based on the second-order stochastic dominance rule, the higher moments of return series, the Shannon entropy, and some other actual investment constraints, we construct a multiconstraint portfolio optimization model, aiming at comprehensively weighting the returns and risk of portfolios rather than blindly maximizing its returns. Furthermore, the whale optimization algorithm based on FTSE100 index data is used to optimize the above multiconstraint portfolio optimization model, which significantly improves the rate of return of the simple diversified buy-and-hold strategy or the FTSE100 index. Furthermore, extensive experiments validate the superiority of the whale optimization algorithm over the other four swarm intelligence optimization algorithms (gray wolf optimizer, fruit fly optimization algorithm, particle swarm optimization, and firefly algorithm) through various indicators of the results, especially under harsh constraints.

This paper presents conceptual comparison among spotted hyena optimiser, grey wolf optimiser, particle swarm optimisation, ant colony optimisation, gravitational search algorithm, bat algorithm, moth flame optimisation, and whale optimisation algorithm. The behaviour of these algorithms is mathematical modelled to show the optimisation process. Twenty-three benchmark test functions are used to validate the performance of these algorithms. The exploration and exploitation of these algorithms are analysed using convergence curve. The experimental results depict that spotted hyena optimiser and grey wolf optimiser give optimal solutions as compared to the other algorithms. Furthermore, these algorithms are tested on five constrained engineering design problems. Experimental results reveal the applicability of these algorithms in real-life engineering design problems.

SummaryThe search for an effective nature‐inspired optimization technique has certainly continued for decades. This work proposes a novel robust multi‐user detection algorithm based on Grey wolf optimization and differential evolution algorithm to overcome the problem of high bit error rate (BER) in multi‐user detection under an impulse noise environment. The simulation results show that the iteration times of the multi‐user detector based on the proposed algorithm is less than those of genetic algorithm, differential evolution algorithm, Grey wolf optimization algorithm, salp swarm algorithm, grasshopper optimisation algorithm, and whale optimization algorithm with the lowerst BER value.