Bio-inspired Computing: Theories and Applications

The forecasting of carbon emissions trading market price is the basis for improving risk management in the carbon trading market and strengthening the enthusiasm of market participants. This paper will apply machine learning methods to forecast the price of China’s carbon trading market. Firstly, the daily average transaction prices of the carbon trading market in Hubei and Shenzhen are collected, and these data are preprocessed by PCAF approach. Secondly, a prediction model based on Radical Basis Function (RBF) neural network is established and it parameters are optimized by Particle Swarm Optimization (PSO). Finally, the PSO-RBF model is validated by actual data and proved that the PSO-RBF model has better prediction effect than BP or RBF neural network in China’s carbon prices prediction, indicating that it has more significant rationality and applicability and deserves further popularization.

In multi-agent system (MAS), the coalition formation (CF) is an important problem focusing on allocating agents to different tasks. In this paper, the single-task single-coalition (STSC) formation problem is considered. The mathematical model of the STSC problem is built with the objective of minimizing the total cost with the ability constraint. Besides, an efficient restart-enhanced genetic algorithm (REGA) is designed to solve the STSC problem. Furthermore, this paper constructs a comparison experiment, employing a random sampling method, an estimation of distribution algorithm and a genetic algorithm without restart strategy as competitors. The results of statistical analysis by the Wilcoxon’s rank-sum test demonstrate that the designed REGA performs better than its competitors in solving the STSC cases of different scales.

The Non-dominated Sorting Genetic Algorithm III (NSGA-III) uses a niche selection strategy based on reference points to maintain the population diversity. However, in an evolutionary process, areas near certain reference points which have no solution attached cannot be searched. To ensure the algorithm searching the entire solution space, and in particular, to avoid some areas not being explored due to no solution existing in the regions currently, we propose a uniform pool reservation strategy based on reference points in this paper. The strategy uses the individuals which are the closest to each reference point to guarantee population diversity. The improved algorithm is compared with classical algorithms based on decomposition and other improved algorithms based on NSGA-III respectively. The performance of each algorithm is evaluated by using inverted generational distance (IGD) and spread. The experimental results show the performance of the improved algorithm.

With the development of China’s economy, more and more energy consumption has led to increasingly serious environmental problems. Faced with the enormous pressure of large amounts of carbon dioxide ( $$\mathrm{CO}_2$$ ) emissions, China is now actively implementing development strategy of low-carbon and emission reduction. Through the analysis of the influencing factors of $$\mathrm{CO}_2$$ emissions in China, five key influencing factors are selected: urbanization level, gross domestic product (GDP) of secondary industry, thermal power generation, real GDP per capital and energy consumption per unit of GDP. This paper applies the Elman neural network optimized by the firefly algorithm (FA) to forecasting the $$\mathrm{CO}_2$$ emissions in China. And the results show that the performance of the FA-Elman is better than the Elman and BPNN, verifying the effectiveness of the FA-Elman model for the $$\mathrm{CO}_2$$ emissions prediction. Finally, we make some suggestions for low-carbon and emission reduction in China by analyzing key influencing factors and forecasting $$\mathrm{CO}_2$$ emissions using FA-Elman from 2017 to 2020.

As air pollution becomes increasingly critical, “near-zero emission” technological innovation in coal-fired plants are needed for the government and public consumers. The aim of this paper is to built the evolutionary game for analysing “near-zero emission” technological innovation diffusion in coal-fired plants. According to bionics research of evolution, this paper introduces the co-evolutionary algorithm to simulate the diffusion. By modeling the evolutionary gaming behavior of coal-fired plants, the simulation can capture the dynamics of coal-fired plants’ strategy, which is adopting “near-zero emission” technological innovation or not. It is key to model the diffusion under electricity market and government regulation because it can provide some suggestions for promoting the diffusion. Simulations show that the coal-fired plant for most profit should adopt independent R&D for “near-zero emission” technology and increasing the subsidy intensity has a significant role in promoting the diffusion.

Amplitude Variation with Offset (AVO) elastic parameter inversion is a nonlinear optimization problem. When a linear or quasi-linear method is used to solve the problem, the inversion result will be unreliable or inaccurate. In this paper, the immune clonal selection algorithm is applied to the AVO elastic parameter inversion problem. The algorithm adopts the specific initialization strategy from Aki’s and Rechard’s approximation equation used in the elastic parameter inversion process to smooth the initialization parameter curve. Additionally, the genetic operation in the algorithm is accordingly improved. A large number of experiments show that this method can significantly improve inversion accuracy.

The traditional pests image recognition technology is based on the point features and line features of the image. In the case of complex lighting conditions or changing camera angles, the classification recognition effect is inaccurate. This article proposes a pests image classification method based on improved Wolf Pack Algorithm (WPA) to optimize Bayesian Network (BN) structure learning. Firstly, We select a pre-trained Convolutional Neural Network (CNN) to extract the image features of data set. And then input the feature vectors and classification of images into BN. Secondly, improved the traditional Wolf Pack Algorithm and used as a search algorithm, Bayesian Information Criterion (BIC) as a scoring function to learn the structure of BN. Then the parameters of BN are learned by Maximum Likelihood (ML) algorithm to form a Bayesian Classifier. Compared with other pest classification method, this method has a certain extent improvement in the classification accuracy of pest image classification.

Cooperative co-evolution (CC) is a promising method for large-scale optimization problems. The performance of a CC framework is affected greatly by variable grouping. DG2 proposed by Omidvar et al. is accurate for variable grouping. DG2, which is a parameter-free differential grouping method, can distinguish overlapping components of decision variables of a function. In this paper, we test DG2 on high-dimensional functions with more than 1000 dimensions. We also test DG2 on the functions that it is highly imbalanced that the contribution of different components to the overall fitness. The experimental results show that the performance of DG2 is stable as the increase of the dimensionality of the functions, but the grouping accuracy of DG2 drops when the imbalance of contribution of different components to the overall fitness becomes greater and greater.

Membrane computing models based on cell structure and function have important applications in computer science and provide new theories and methods for modeling biological systems. The spiking neural P system based on the min-sequentiality strategy is a special kind of membrane computing model. Anti-spikes and inhibitory functions are introduced into spiking neural P systems based on the min-sequentiality strategy. We construct two sequential spiking neural P systems with anti-spikes in different ways. The corresponding modules of the two systems are designed separately. Finally, we prove that the two spiking neural P systems with anti-spikes based on the min-sequentiality strategy are universal as both number generators and acceptors.

Molecular computing [1] is a field with a great potential and fastest growing area of Computer Science. Although some approaches to solve NP hard problems were successfully accomplished on DNA strand, only few results of practical use so far. A direction of molecular computing namely Gene assembly in ciliates has been studied actively [3] for a decade. In the present paper, we use a variant of gene assembly computing model of Guided recombination system with only two operations of insertion and deletion [7] as a decision problem solver. We present our results of parallel algorithms which solve computational hard problems HPP and CSP, in an efficient time.

How to optimize industrial structure to meet coordinated development of economy, society and environment has always been a key issue in research and management. In this paper, an optimization model based on MOEA is proposed to adjust industrial structure to meet the increasing demand. Increasing economy and employment along with reducing carbon emission are the objectives of model, which is solved by one of MOEA, NSGA-II. Jiangsu-Zhejiang-Shanghai is studied as the case. The Pareto fronts of solutions show good convergence and robustness. The optimization methods are compared with each other from operation efficiency and significance which turns out NSAG-II has advantages in studying this issue. Results are also analyzed in different perspective and discussed under Flexible Optimization. The idea of applying MOEA in industrial structure optimization provides a scientific way to promote economic growth and employment along with answering green call for low-carbon life.

DNA strand displacement is widely used in the construction of DNA molecule computational models. In this work, nicking enzyme is used as the input of the logic calculation model for it can cut one strand of a double-stranded DNA at a specific recognition nucleotide sequences known as a restriction site. Based on this, a variety of logic gates are designed and implemented, and a multi-person voting circuit is constructed.

Electrocephalogram(EEG) signals classification is an important problem in the field of brain computer interface. There are many EEG signals classification methods, but most of are not very efficient in this problem. Deep learning had been broadly used in image classification and has significant performance in classifying images. This paper proposes a comprehensive spatio-temporal feature classification method based on deep learning. It combines Convolutional Neural Network (CNN) and Long-term Short-term Memory network (LSTM) to the motor imaginary EEG classification. Experimental results show that it can preserve spatial, frequency and temporal features of motor imaginary EEG simultaneously and improves the classification accuracy of EEG signals.

The satisfiability problem (SAT) is one of the NP-complete problems in the fields of theoretical computer and artificial intelligence, is the core of NP-complete problems. Compared with traditional DNA self-assembly, DNA origami is a new method of DNA self-assembly. We first give a description and the status quo of study of the satisfiability problem, briefly introduce the principle of DNA origami, propose the computing model based on DNA origami to solve the satisfiability problem, and solve an instance with 3 variables, 3 clauses to illustrate the feasibility of the algorithm. The proposed model only uses gel electrophoresis to search the solution to the problem, which is the most reliable biological operation known to date, therefore the proposed model is feasible. At present, the reported results concerning using origami to solve the NP-complete problem is relatively few. Our method is a new attempt to solve the NP- complete problem using biological DNA molecules.

Self-assembly reveals the essence of DNA computing, DNA self-assembly is thought to be the best way to make DNA computing transform into computer chip. This paper introduce a method of DNA 3D self-assembly algorithm to solve the Maximum Clique Problem. Firstly, we introduce a non-deterministic algorithm. Then, according to the algorithm we design the types of DNA tiles which the computation needs. Lastly, we demonstrate the self-assembly process and the experimental methods which could get the final result. The computation time is linear, and the number of the distinctive tile types is constant.

In this paper, a deep neural network model is proposed to predict industrial air pollution, such as PM2.5 and PM10. The deep neural network model contains 9 hidden layers, each layer contains 45 neurons. The output of the hidden layer neurons is calculated using the ReLU activation function, which can effectively reduce the gradient elimination effect of the deep neural network. Twelve air pollutant indicators from industrial factories are collected as the input data, such as CO, NO2, O3, and SO2. About 180,000 real industrial air pollution data from Wuhan City are used to train and test the DNN model. Furthermore, the performance of our approach is compared with the SVM and Artificial neural network methods, and the comparison result shows that our algorithm is accurate and competitive with higher prediction accuracy and generalization ability.

Finding a constraint shortest path which passes through a set of specified vertices is very important for many research areas, such as intelligent transportation systems, emergency rescue, and military planning. In this paper, we propose an efficient genetic algorithm for solving the constraint shortest path problem. Firstly, the Dijkstra algorithm is used to calculate the shortest distance between any two specified vertices. The optimal solution change from the original problem into the Hamilton path problem with the specified vertices. Because the number of specified vertices is much less than the number of vertices for the whole road network, the search space would be reduced exponentially. Secondly, the genetic algorithm is adopted to search for the optimal solution of the Hamilton path problem. Thirdly our algorithm should detect and eliminate the cycle path. Finally, the performance of our algorithm is evaluated by some real-life city road networks and some randomly generated road networks. The computational results show that our algorithm can find the constraint shortest path efficiently and effectively.

Attribute reduction is an important issue in rough set theory. Many heuristic algorithms have been proposed to compute the minimal attribute reduction since it is a NP hard problem, while most of them have the drawback to fall into local optimal solution. The other way to solve this problem is based on parallel computing. Membrane computing model is a distributed, maximal parallel and non-deterministic computing model inspired from cell. In this paper, we attempt to solve the attribute reduction problem by membrane computing, and propose a cell-like P system $$\varPi _{AR}$$ to compute all exact minimal attribute reductions with $$\mathrm{O}(m \log n)$$ time complexity.

In this paper, we quantify the interactive pattern between two time series: the number of users (NoU) representing user’s activity (UA) and downlink traffic load (DTL) generated from the base station (BS). We model the characteristics of UA, and use K-means clustering algorithm to characterize the hidden spatial association pattern in the wireless cellular system. The results show that (1) there is a strong linear interaction between UA and DTL; (2) the NoU has a strong weekdays and weekends mode. (3) the results of clustering well match the reference scenario information, with the scenario recognition accuracy of 75%. We demonstrate that such approach proposed can identify the scenario of the BSes, which can help us understand the spatial temporal traffic patterns of wireless cellular system.

Membrane computing is a computational model abstracted from the structure and function of biological cells. Since membrane computing system (also known as P system) was proposed, researchers designed many P systems and P system simulators. However, because of the diversity of evolutionary rules, it is difficult to find suitable simulation tools to implement these P systems. Based on the cell-like P system, this paper proposed a universal P system description language (called UPL) and a universal P system simulator (called UPS). UPL supports the expansion of membrane structural characteristics and the combination of various rule types. UPS can simulate the P system described by UPL. The experimental results verify their effectiveness.

Dynamic multimodal optimization (DMO) problem is introduced and solved with brain storm optimization (BSO) algorithms in this paper. A dynamic multimodal optimization problem is defined as an optimization problem with multiple global optima and characteristics of global optima are changed during the search process. The effectiveness of BSO algorithm is validated on a test problem which was constructed based on the dynamic optimization and multimodal optimization. Results show that BSO algorithm is an efficient and robust optimization method for solving dynamic multimodal optimization problems.

In MOEA/D, the replacement strategy plays a key role in balancing diversity and convergence. However, existing adaptive replacement strategies either focus on neighborhood or global replacement strategy, which may have no obvious effects on balance of diversity and convergence in tackling complicated MOPs. In order to overcome this shortcoming, we propose a hybrid mechanism balancing neighborhood and global replacement strategy. In this mechanism, a probability threshold $$ p_{t} $$ is applied to determine whether to execute a neighborhood or global replacement strategy, which could balance diversity and convergence. Furthermore, we design an offspring generation method to generate the high-quality solution for each subproblem, which can ease mismatch between subproblems and solutions. Based on the classic MOEA/D, we design a new algorithm framework, called MOEA/D-HRS. Compared with other state-of-the-art MOEAs, experimental results show that the proposed algorithm obtains the best performance.

Many memristor-based neural network arrays that have been proposed in recent years are simultaneously dealt with all of their signal inputs in signal reception status. Therefore, when a relatively small-scale neural network is implemented with this memristive array, some of the inputs which are not used may cause errors in the result due to the impact of an unexpected signal. In this paper, a flexible memristor-based neural network is proposed. Based on this network, the number of synapses used at work can be flexibly configured according to the required size, thereby improving system performance. The memristor-based neural network is simulated in Pspice to implement two different scales, which proves the feasibility and effectiveness of a flexible memristive neural network.

Job shop scheduling problem (JSP) is a well-known combinatorial optimization problem of practical importance, but existing evolutionary algorithms for JSP often face problems of low convergence speed and/or premature convergence. For efficiently solving JSP, this paper proposes a memetic algorithm based on biogeography-based optimization (BBO), named BBMA, which redefines the migration and mutation operators of BBO for JSP, employs a local population topology to suppress premature convergence, and uses a critical-path-based local search operator to enhance the exploitation ability. Numerical experiments on a set of JSP instances show that the proposed BBMA has significantly performance advantage over a number of state-of-the-art evolutionary algorithms.

A novel chaotic system of three-dimensional mathematic model is proposed in this paper. According to the changes of system parameters and initial values, the dynamical behaviors of the system are investigated in detail by using the classical dynamical analysis methods, such as Lyapunov exponents, bifurcation diagrams etc. Some abundant dynamical phenomena, such as chaos, transient chaos and period-doubling and so on are observed in numerical simulation by Matlab. The simulation results of Matlab can further prove the feasibility and flexibility of this system.

Flow-shop scheduling problem (FSP) is a well-known NP-hard combinatorial optimization problem that occurs in many practical applications. Traditional algorithms are only capable of solving small-size FSP instances, and thus many metaheuristic algorithms have been proposed for efficiently solving large-size instances. However, most existing algorithms still suffer from low convergence speed and/or premature convergence. In this paper, we propose an enhanced biogeography-based optimization (BBO) algorithm framework for FSP, which uses the largest ranked value representation for solution encoding, employs the NEH method to improve the initial population, and designs a reinsertion local search operator based on the job with the longest waiting time (JLWT) to enhance exploitation ability. We respectively use the original BBO migration, blended migration, hybrid BBO and DE migration, and ecogeography-based migration to implement the framework. Experimental results on test instances demonstrate the effectiveness of the proposed BBO algorithms, among which the ecogeography-based optimization (EBO) algorithm version exhibits the best performance.

There is increasing evidence that long non-coding RNAs (lncRNAs) are closely related to many human diseases. Developing powerful computational models for potential lncRNA-disease association identification would facilitate biomarker identification and drug discovery for human disease diagnosis, treatment, prognosis and prevention. Now there exist a number of methods specially for this problem based on inductive matrix completion, random walk or classification. In terms of this issue, classification has just come to the fore. Extracting important features from disease network and RNA network, namely network embedding, is the top priority. Moreover, taking the complexity into consideration, genetic algorithm is adopted to tune the hyper-parameters of our network embedding model. Due to a lack of negative samples, we also exploit Positive-Unlabeled (PU) learning to help out. In brief, we propose a weighted bagging lightGBM model for lncRNA-disease association prediction based on network embedding and PU learning.

Android is the most popular smartphone operating system in the world thanks to its openness, which also attracts many Android malware writers. It is really a big challenge for the various Android markets to filter out malware accurately and quickly before provisioning a large number of APPs. Many handcraft feature-based detection solutions had been proposed for solving this problem. But the malware writers can always find ways to change the features while maintaining the malware’ malicious semantic. Inspired by the findings in biology, we advocate identifying Android APPs’ genes that are responsible for the malicious behaviors. Based on this idea, we proposed a new method called DroidGene, which treats calling sequences and permissions as DNA, and using elaborately designed LSTM to find APPs’ malicious genes. The result of experiments on 16,200 Android samples shows that both the accuracy (99.1%) and the detection time (0.36 s) of DroidGene are superior to the state-of-the-art method.

Logo recognition by Convolutional Neural Networks (CNNs) on a smartphone requires the network to be both accurate and small. In our previous work [1], we proposed the accompanying dataset method for single logo recognition to increase the recall and precision of the target logo recognition. However, the reason why it works was unclear, thus it was hard to compress the network while maintaining the same accuracy. In this paper, we use DeconvNet [9] to visualize our network’s feature maps and propose a metric to analyze them quantitatively. Finally, we obtain a better understanding of the influences in the network brought by accompanying datasets. Under its guidance, an effective way to compress the network is devised by us. The experiments show that we can reduce the size of the neural network’s first layer by 30% while only lower the recall and precision by 0.014 and 0.01. The training time is also saved by 40% due to the network compression.

Artificial neural networks (ANNs) have powerful function approximation and pattern classification capabilities, but their performance is greatly affected by structural design and parameter selection. Traditional training methods have drawbacks including long training time, over-fitting, premature convergence, etc. Evolutionary optimization algorithms have provided an effective tool for ANN parameter optimization, but simultaneously optimizing ANN structure and parameters remains a difficult problem. This paper adapts a relatively new evolutionary algorithm, water wave optimization (WWO), for both structure design and parameter selection for ANNs. The algorithm uses a variable-dimensional solution representation, and designs new propagation, refraction, and breaking operators to effectively evolve solutions towards the optimum or near-optima. Computational experiments show that the WWO algorithm exhibits significant performance advantages over other popular evolutionary algorithms including genetic algorithm, particle swarm optimization, and biogeography-based optimization, for ANN structure and parameter optimization.

This paper proposes an adaptive method to select recombination operators, including differential evolution (DE) operators and polynomial operators. Moreover, a push and pull search (PPS) method is used to handle constrained single-objective optimization problems (CSOPs). The PPS has two search stages—the push stage and the pull stage. In the push stage, a CSOP is optimized without considering constraints. In the pull stage, the CSOP is optimized with an improved epsilon constraint-handling method. In this paper, twenty-eight CSOPs are used to test the performance of the proposed adaptive GA with the PPS method (AGA-PPS). AGA-PPS is compared with three other differential evolution algorithms, including LSHADE44+IDE, LSHADE44 and UDE. The experimental results indicate that the proposed AGA-PPS is significantly better than other compared algorithms on the twenty-eight CSOPsq.

Port scanning is a widely used technology in reconnaissance, which aims to determine remotely the running services on the target TCP/UDP ports. Current research works have achieved acceptable performance for detection of conventional port scanning, which use handcrafted features such as packets receiving rate, count of requesting ports and packets arriving time distribution. However, advanced attacks such as APT usually employ low-speed scans to lower the risk of exposure. Nevertheless, it is a challenge to precisely detect a low-speed scan since it has much coarser features that are hard to be matched by the current approaches. We propose a novel method DeepPort to solve this problem. DeepPort filters out a majority of normal packets using their well-defined features. Thereafter, DeepPort detects port scans using learned features using a dedicated Convolutional Neural Network (CNN) that is trained from real scanning packets under various time interval configurations. The experiments carried in our campus network show that DeepPort can detect 10 class of low-speed scans with a precision of 97.4% and a recall of 96.9%.

The decomposition-based algorithms, such as MOEA/D, transform a multiobjective optimization problem into a number of single-objective optimization subproblems and solve them in a collaborative manner. It is a natural framework for using single-objective local search for solving combinatorial multiobjective optimization problems. However, the performance of the decomposition-based algorithms strongly depends on the shape of PFs. For this purpose, this paper proposed a dual-population-based local search in MOEA/D framework (DP-MOEA/D-LS) to address the multiobjective traveling salesman problem. Two populations using different sets of direction vectors and different decomposition approaches cooperate with each other for achieving appropriate balance between the convergence and diversity. The experimental results show that DP-MOEA/D-LS significantly outperforms the compared algorithms (MOEA/D-LS (WS, TCH, PBI and iPBI)) on all the test instances.

The effectiveness of most of the existing decomposition-based multi-objective evolutionary algorithms (MOEAs) is yet to be heightened for many-objective optimization problems (MaOPs). In this paper, a cone decomposition evolutionary algorithm (CDEA) is proposed to extend decomposition-based MOEAs to MaOPs more effectively. In CDEA, a cone decomposition strategy is introduced to overcome potential troubles in decomposition-based MOEAs by decomposing a MaOP into several subproblems and associating each of them with a unique cone subregion. Then, a scalarization approach of adaptive direction penalized distance is designed to emphasize boundary subproblems and guarantee the full spread of the final obtained front. The proposed algorithm is compared with three decomposition-based MOEAs on unconstrained benchmark MaOPs with 5 to 10 objectives. Empirical results demonstrate the superior solution quality of CDEA.

Most of existing swarm intelligence (SI) algorithms is modeling based on natural phenomena. Firstly, different from the previous practices, this paper constructs a mathematical model based on the traditional optimization algorithms. To simplify this model, a new algorithm Linear Transformation and Elitist Selection algorithm (LTES) is proposed. Experiment shows that the algorithm has origin illusion phenomenon. Then, this paper observes origin illusion phenomenon for the population-based optimization algorithm, and experiments shows that crossover operator is an effective way for LTES’ origin illusion problem. Finally, another algorithm Contraction and Guidance Algorithm (CGA) is proposed to prove that elitist selection is not necessary. The experimental results show that both algorithms are effective.

This paper addresses the Vehicle Routing Problem with Service Time Customization (VRPTW-STC), which is an extension of the classic Vehicle Routing Problem with Time Window (VRPTW). In VRPTW-STC, the decision maker tries to find an optimum solution with the smallest fleet size, the lowest travelling distance as well as the largest total service time of all customers. The objective to enlarge each customer’s service time obviously conflicts with the need of reducing both changeable and fixed transport costs, i.e. travelling distance and fleet size. At the same time, the routing plan must meet the time window constraint and the vehicle capacity constraint. To solve this problem, we designed a Multi Ant System (MAS) based hybrid heuristic algorithm inspired by to decompose a multi-objective problem into several single objective ones. Then, Ant Colony Optimization (ACO) algorithms are applied to every single-objective problem. A unique global best solution is maintained to record the current best solution. The global best solution will be updated when a new feasible solution found by any ACO dominate current global best solution. Several local search algorithms are also incorporated into MAS to help improve the solution quality. Solomon’s benchmark tests are used to test the effectiveness of the proposed algorithm. The computation experiment results show that our proposed MAS based hybrid heuristic algorithm performs better than typical existing algorithms.

This paper presents the MPC (Model Predictive Control) method based on CFD (Computational Fluid Dynamics), aiming to optimize the temperature control of the data center. The paper establishes the three-dimensional physical model of the data center according to the boundary conditions, gets the unit step function response of the input and output temperature by the steady and unsteady simulation solution, then gets the mathematical model of data center temperature by system identification. The MPC simulation experiment is carried out, compared with the traditional PID control, resulting in that MPC has better control quality and has great application values on the temperature control of the data center.

Artificial Intelligent Systems have shown great potential in the musical domain. One task in which these techniques have shown special promise is in the automatic music composition. This article describes the development of an algorithm for designing musical expressiveness for a tonal melody generated by computer. The method employed is based on a model of self-recognition of the harmonic structures contained in the melody and, by means of the “harmonic function” carried by every single one of these, provides useful information for the dynamics. The article is intended to demonstrate the effectiveness of the method by applying it to some (tonal) musical pieces of the 18th and of the 19th century. At the same time it is going to indicate ways to improve the method.

Satellite is an important space platform today. Achieving reasonable satellite management control greatly affects the development of the aerospace field. Multi-satellite and multi-station mission planning system was proposed to achieve control of satellite resources. In the system, planning algorithms are particularly important, so we proposed a mathematical model for the mission planning of multi-satellite and multi-ground station. Then, we proposed a hybrid dynamic population genetic algorithm (HDPGA) for satellite mission planning. In this algorithm, the large-population size is used for global optimization and the small-population size is used for local improvements. Additionally, the mission planning algorithm (MPA) is used to arrange the mission sequence on the ground station time window. We designed multiple sets of experiments to verify the effect of HDPGA. The results show that our proposed algorithm can meet the needs of the planning system. At the same time, HDPGA is better than the other four algorithms.

DNA strand displacement is a recently emerging technology which is widely used in constructing logic circuit, building a bio-chemical computer and etc. A bio-chemical 8 to 3 priority encoder based on DNA strand displacement technology which is the basic unit of a bio-chemical computer is introduced in this paper. The proposed encoder is verified by the simulation using programming language Visual DSD. Simulation results further confirm that the DNA strand displacement technology is a promising method which could be implemented in bio-chemical circuit and computers.

In this paper, a biological sensing model based on graphene oxide is proposed. A YES gate and a AND gate are constructed by using the ability of graphene oxide to adsorb single strand and quench fluorescence. The biosensor we designed can be used not only as a logical element, but also to detect a specific target DNA. Then, taking YES gate as an example, orthogonal experiments, condition optimization and target selective detection are carried out to demonstrate the practical significance of the sensing model designed. In subsequent experiments, we will design more complex logic components on this basis and try to apply them to practice.

Medium and long-term power load forecasting is the basis for power system planning and construction. This paper builds a prediction model based on SaDE-SVM algorithm. In order to reduce its selection problem of excessive large-scale hyperplane parameters, improve global optimization ability of traditional SVM, and further improve the prediction accuracy of SVM, the SaDE-SVM optimization algorithm is proposed. This algorithm optimizes the training process of traditional SVM based on adaptive differential evolution algorithm. The results of the medium and long-term forecasting for China’s power load show that the improved SaDE-SVM algorithm has good adaptability, robustness, fast convergence rate, and high accuracy for multi-influencing factors prediction model with less data volume, and is applicable to relevant medium and long-term forecasts.

Establishing a scientific and reasonable mid- and long-term power load forecasting method is the premise of power industry planning and construction. This paper constructs a hybrid electric load forecasting model based on Gaussian process (GPR) and particle swarm optimization (PSO). The paper uses the PSO algorithm to optimize the parameters in the co-variance function, and uses the modified parameters as the initial value to train the power load in the GPR model. Under the Bayesian framework, the parameters in the co-variance function are again optimized. Finally, the trained GPR model is used to predict the power load, and the results are compared with the auto-regressive integral moving average model and the exponential smoothing model. The verification results show that the hybrid electric load forecasting model based on Gaussian process (GPR) and particle swarm optimization (PSO) has good stability and higher prediction accuracy, and is suitable for medium and long-term electric load forecasting.

Feature selection is a vital step in many machine learning and data mining tasks. Feature selection can reduce the dimensionality, speed up the learning process, and improve the performance of the learning models. Most of the existing feature selection methods try to find the best feature subset according to a pre-defined feature evaluation criterion. However, in many real-world datasets, there may exist many global or local optimal feature subsets, especially in the high-dimensional datasets. Classical feature selection methods can only obtain one optimal feature subset in a run of the algorithm and they cannot locate multiple optimal solutions. Therefore, this paper considers feature selection as a multimodal optimization problem and proposes a novel feature selection method which integrates the barebones particle swarm optimization (BBPSO) and a neighborhood search strategy. BBPSO is a simple but powerful variant of PSO. The neighborhood search strategy can form several steady sub-swarms in the population and each sub-swarm aims at finding one optimal feature subset. The proposed approach is compared with four PSO based feature selection methods on eight UCI datasets. Experimental results show that the proposed approach can produce superior feature subsets over the comparative methods.