Search

In the present work, an algorithm for calibrating a 2D camera as a tool for a 6-axis robot is presented. The aim of the present development is to introduce a new, faster, precise, and more reliable way to calibrate a camera as a robot tool. From the proposed algorithm optimization of the time, it takes to calibrate the camera and start programming is achieved. The time is decreased by 80% and the accuracy and reliability of the system are increased.

This article mainly optimizes the positioning error, positioning optimal value and positioning adaptability of the optimal positioning algorithm of navigation and positioning equipment based on the multi-navigation model, and improves the accuracy of the positioning optimal value through algorithm optimization. In order to ensure the effectiveness of the proposed improved algorithm, by designing simulation experiments, conducting data comparison tests on it, and drawing experimental conclusions based on the data comparison results.

To solve the Autonomous Vehicle (AV) problem, an artificial learning model that translates sensor data into controls must be built. This way, the vehicle can react to its changing environment. To the best of our knowledge, there are no platforms where researchers can both develop new machine learning models, train them and compare them directly to others. We believe that such a platform can greatly impact the field of artificial intelligence and AV. In this paper, we propose an enhanced version of the Learning Algorithm Optimization Platform LAOP, called LAOP 2.0. It helps researchers in the field of artificial intelligence develop their models by allowing them to easily test and compare them. We also introduce the Learning Algorithm Sharing Platform (LASP), which makes it easy to share deep learning algorithms. As a demonstration of the versatility of our platform, we compared two learning algorithms. The first one, Fully Connected Neural Network (FUCONN), uses reinforcement learning to train itself. The second one, Mimicking HUman Behaviour (MHUB), uses supervised learning to adjust its weights and learns from human input. We demonstrate through extensive simulations that FUCONN outperforms MHUB after being trained.

This paper mainly studies the modeling of classical processing workshops with the aim of manufacturing period and no recycling. Based on the optimization of production scheduling of printed circuit boards, an optimization algorithm is proposed. Firstly, aiming at minimizing the work penalty, the integer programming model of the printed circuit board of the single flow shop is established. Then, under the premise of chain priority constraint, by combining the minimum relaxation priority distribution rule and the branch and bound method, a heuristic algorithm with dynamic bottleneck transformation with penalty weights is proposed. Finally, through the simulation experiments of data sets, it is verified that the proposed model and algorithm have certain effects in solving the production scheduling problem of printed circuit boards.

With the rapid development of intelligent buildings, the research and development of VAV air-conditioning systems have become the focus of attention. This article mainly studies the temperature control loop of the VAV air-conditioning room, and optimizes the control effect of the temperature control loop of the air-conditioning room by combining intelligent algorithm and PID control. At the same time, a fuzzy PID controller is designed. Fuzzy PID control is added to the temperature control loop of the air-conditioning room. Matlab/Simulink is used for modeling and simulation. The experimental results show that the fuzzy PID control effectively improves the stability of the indoor temperature. The resulting maximum dynamic deviation is smaller, the adjustment time is shorter, and the interference resistance is better, which has certain theoretical significance for improving the comfort of the VAV system air-conditioned room.

Rate distortion optimization (RDO) is the basis for algorithm optimization in video coding [1], such as mode decision, rate control and etc. Minimizing the rate distortion coding cost is usually employed to determine the optimal coding parameters such as quantization level, coding mode, and etc. However, rate and distortion calculations for optimal solution decision from massive possible candidates suffer from dramatically high computation complexity. To resolve this problem, this paper proposes a fast TU level rate model with higher accuracy by fully imitating the behavior pattern hid in entropy.

Recently, Problem-based learning has been increasingly used, and with those practices, self and peer evaluation tasks have gained more importance. This method of learning assists learners in the acquisition of skills and competencies. In order to support self and peer evaluation tasks, there are some different software offers in the market. The ease use of them is demanded since these tasks are an important part of evaluation. With the literature emphasis on the importance of assessment and critical spirit, it arises the possibility to implement formulas to increase the efficiency on the methodology applied and satisfaction of students and teachers. This paper introduces formulation problems about marks weighting in self and peer evaluation. It presents the advantages and disadvantages of different solutions for the formulation and suggestions of adapting them among tools, feedback surveys, and teacher’s methods of evaluation. It is also expected that the formulation, which was implemented in the software tool WEBAVALIA, has a great impact on feedback, improving user experience, and both efficiency and flexibility.

Based on the correlation between precipitation phenomena and different meteorological factors, this paper proposes a new optimization algorithm for raindrop spectrum precipitation phenomena instruments applied to national meteorological observation. Raindrop spectrum precipitation phenomena instruments has been formally applied used in China meteorological observation stations for more than two years. Due to the influence of external environment,working principle, and algorithm limitations, obvious misidentification sometimes may occur, and the identification accuracy still needs to be improved. At this stage, it is still necessary to rely on manual observation to correct it, which has become a constraint affecting the automation reform of ground meteorological observation in China. The working principle of the raindrop spectrum precipitation phenomena instruments is based on the measured precipitation particle size and velocity, and the precipitation phenomena is recognized in the Atlas-Ulbrich curve distribution, and combined with air temperature, humidity and visibility factors for quality control, which has its limitations and one-sidedness, a certain number of misidentification data existed in meteorological observation. Precipitation phenomena is a systematic and diverse weather change process, data analysis display it is closely related to air temperature, humidity, visibility, air pressure, wind speed, precipitation,sunshine, ground temperature and cloud, etc. The occurrence of precipitation phenomena corresponds to the change or not changer of relevant meteorological factors. Statistics show that different precipitation phenomena need to meet different meteorological conditions. For example, the meteorological condition of drizzle is “the stability of each factor is strong, the humidity is high, the wind speed is low, the visibility is poor, the precipitation is little and there is no sunshine.” The meteorological condition of hail is “air temperature, humidity and air pressure obviously change, ground temperature obviously drop, poor visibility, high wind speed, high precipitation intensity, no sunshine.” Therefore, based on the existing algorithm of the raindrop spectrum precipitation phenomena instruments, this paper proposes a new identification optimization algorithm for the raindrop spectrum precipitation phenomena instruments used in meteorological observation, and scientifically and rationally introduces meteorological variables for multi-factor comprehensive identification. The algorithm can greatly improve the accuracy of identification, improve the availability and effectiveness of observation data, reduce manual corrections, full meet the needs of weather forecasting, climate analysis, scientific research and meteorological services.

Medical Question Answering (medical QA), which studies the problem of automatically answering patients' medical questions online, is one of the major applications of bioinformatics. Though many efforts have been made before, the medical QA system still deserves delicate algorithm optimization due to the serious application scenario and strict requirement for the answer quality. In this paper, we introduce a novel Knowledge Abstraction Matching (KAM) method for the medical QA problem. The intuition of KAM is that there are many frequent repeat text segments appearing in the answers across different questions. From this view, we propose a new method that consists of frequent segment $N$ -gram mining, medical knowledge abstraction, medical segment matching and answer re-retrieval. KAM has been incorporated into Baidu's enterprise medical QA system MelodyQA deployed on the backend of Muzhi Doctor. The evaluation shows that the proposed method can generate more quality answers for MelodyQA with a significant improvement of question coverage under acceptable accuracy.

Humans are basically curious creatures. We always look for all the latest information through books or news. News usually contain tags or categories to make it easier to find similar news. It can be done manually by human or with machine to make it easier. Data mining technique such as clustering and classification can be used to help the categorization process. Clustering is a technique that can be used to grouped news based on their similarity. There is some algorithm that can be used to clustering data. One of it is Ant Colony optimization (ACO). ACO is a metaheuristic algorithm that mimics how ants brings food to the colony. ACO also can be used to solve clustering problem. In this paper, we will use the ACO version for clustering to cluster news and find the performance of said method. Then we compare the result with ground truth.

Aiming at the problem of suppression and removal of speckle noise in SAR images, Lee filter is considered to be a relatively good algorithm by comprehensive considerations. It has good performance in many scenes. The main disadvantage is that its calculation takes a long time and the real-time performance is too poor for large-format high-resolution SAR images. Therefore, the GPU devices can be used for acceleration in algorithm optimization. This paper first implements the classic Lee filtering algorithm in Python environment, and then combines Numba to realize the CPU acceleration of the algorithm. Then, the classical Lee filtering algorithm is repeated by C/C++. Finally, the GPU acceleration of the algorithm is implemented by CUDA. By comparison, the results show that GPU acceleration can increase the speed by nearly two orders of magnitude.

Aiming at the problems of large subjectivity and inaccurate diagnosis results in the fault diagnosis of tank gun control system, the fault diagnosis method based on improved artificial bee colony is studied. Combined with the improved artificial bee colony algorithm and BP neural network, a BP neural network algorithm based on improved bee colony optimization algorithm is formed and the model of the algorithm is established. And through the use of MATLAB simulation of computer programs, compared with the BP neural network algorithm without optimization, the experiment is summarized. The results show that the system can give fault diagnosis results more accurately, which helps to improve the maintenance efficiency and reliability of the tank gun control system.

With the development of Internet, information overload about products is pervasive. It is important for commercial platforms to predict users’ preferences and recommend information. Neighborhood-based recommendation algorithm is one of the most popular methods, which are used to predict the rating of items that have not yet been rated. However, neighborhood-based algorithm suffers from data sparsity in practice, causing low accuracy. A recommendation algorithm is proposed in this paper, in which auxiliary information of users and items are utilized. Specifically, clustering method is introduced to divide users into categories, which reduces the time complexity of the algorithm, and latent factor model approach is applied to predict user-item matrix, which improves the accuracy of neighborhood-based algorithm. Experiments on real-world dataset demonstrate that the proposed KC-LFMCF approach is superior to the conventional neighborhood-based algorithm in terms of recommendation accuracy and time complexity.

This paper presents a new hybrid heuristic algorithm for the job shop scheduling problem with maintenance, delay and setup costs, and setup times. A mathematical model was initially developed to solve the problem composed of up to 10 orders. Solving large-scale problems, however, proved to be extremely time-consuming due to existing limitations and variables within the model; therefore, the hybrid simulated annealing (HSA) method was presented for large-scale problems. HSA efficiently solved the large-scale problems with a large number of orders in a proper minimized time.

In order to improve the accuracy of power transformers, a new DGA ratios based on support vector machine (SVM) and quantum-behaved particle swarm optimization with binary encoding (BQPSO) algorithm was proposed in the paper. First of all, 28 DGA ratios are used as the input vectors. Then, the SVM parameters and DGA ratios were simultaneously optimized by BQPSO. Finally, a transformer fault diagnosis model based on the artificial bee colony (ABC) algorithm optimization SVM combined with the cross validation (CV) principle is used to diagnostic fault. The diagnosis results show that the proposed DGA ratios increase the accuracy by 12%-28% over IEC ratios and DGA data. Furthermore, the accuracy of ABCSVM is better than the GASVM and standard SVM (accuracy rate are 84.35% 79.33%% and 65.41%).

For the CNC machine tool, the processing parameters of cutting are a key factor to affect the manufacturing accuracy and tool wear. However, this study proposes a prediction system based on neural network algorithm to estimate the wear of turning tool. For neural network algorithm, the processing parameters, the cutting speed, feed rate and material removal rate are investigated as the input parameters of the BNN. The output parameters of the BNN are the wear of turning tool and the surface accuracy of workpiece. Experimental results showed that the turning cutting wear of prediction accuracy compared with the experiment is 93.44%. The max error of cutting wear between the prediction and the experiment is $15\mu \mathrm {m}.$

This paper mainly proves the feasibility and validity of the overlapping community discovery through the size of attribution. The MB algorithm based on BGLL algorithm is proposed. Firstly, karate data sets are divided based on the traditional community partition algorithm BGLL algorithm. After obtaining the result of partition, the nodes in each sub-community are used to find the corresponding degree of ownership of nodes. The size of each node's membership is then compared to determine which community the node belongs to. Finally, the threshold value is adjusted by modularity to determine whether the nodes under this belongingness are overlapping nodes until the module degree is maximum. The experimental results show that it is feasible to overlap the community nodes with the degree of belonging, and it has certain efficiency of algorithm optimization.

In view of the disadvantages of slow convergence and local optimality in traditional neural network learning algorithms, a neural network nonlinear function fitting system with improved particle swarm optimization (PSO) is designed for intelligent prediction accuracy. The intelligent optimization scheme integrates the immune PSO algorithm with BP neural network theory, and realizes the nonlinear function fitting algorithm of BP neural network, based on IPSO algorithm optimization. The new fitting algorithm first determines the structure of the BP neural network, and optimizes the initial weight and threshold with IPSO algorithm. Then, it improves the intelligent prediction method to verify the performance. Numerical experiments show that the algorithm proposed in this paper improves the fitting ability of BP neural network, reduces the fitting error and improves the fitting accuracy.

A267 at the problem of low accuracy and long training time in the traditional intelligent algorithm of tr ansformer fault diagnosis, a fault diagnosis method based on chaotic bats algorithm optimization fuzzy Petri net (CBA-FPN) is proposed. DGA was considered as feature input, the parameters of FPN were coded into bats, and then the optimal parameters of FPN were founded by chaotic bat optimization algorithm. The fuzzy Petri net model was used to diagnose the transformer according to the optimal parameters. The simulation results show that the proposed algorithm has faster convergence speed and the higher correctness than the method of BP Optimized FPN (BP-FPN) and BA Optimized FPN network(BA-FPN), and has certain application valu e in the transformer fault diagnosis

Computationally complex algorithms are very useful when attempting to accurately predict the secondary structure of RNA. This paper seeks to use rule-based permutation with OpenCL to output every possible RNA sequence that fits a particular protein amino acid sequence (rule).In previous research, two algorithms were explored along with their possible application in the field of bioinformatics and biochemical engineering. One of these two algorithms is further optimized using OpenCL, a cross platform programming language known for its parallel computing potential and its ability to leverage processing power of GPU's and like devices. We propose that improving RNA secondary structure prediction accuracy can be achieved by extensive analysis of natural occurring RNA folds in comparison to others like them that belong to the same Amino acid sequence

Researched a new class of parameters such as NTRU algorithm with optimized basic algorithmic operations. The results obtained with regard to the performance for the NTRU method with the modified parameters were compared with the speed characteristics of the NTRU implementation of the method with the standardized parameters.

Collaborative filtering (CF) is the most popular recommendation algorithm, which predicts unknown ratings through historic data. However, the traditional collaborative filtering algorithm only considers the individual user's historical behavior of items, without considering the users' preferences and attributes. In this paper, a collaborative filtering algorithm based on latent factor model clustering (CF-LFMC) is proposed. Firstly, the user's score matrix is decomposed into two matrices by the latent factor model (LFM) and gets the user-class matrix. Secondly, the user-class matrix is decomposed into many clusters by the clustering algorithm, which presents users' hidden attributes. Finally, based on users' hidden attributes and users' history score record, the similarity of items is calculated and item scores can be predicted. Through experiments on the MovieLens dataset, it shows that the proposed algorithm can reduce the mean absolute error and improve the accuracy of recommendation.

With the rapid technology development of systems that assist drivers, the lack of uniform support for efficient algorithm development has emerged. We propose a universal Software Development Kit (SDK) concept with common optimization routines that can be applied to typical Advanced Driver Assistance Systems (ADAS) algorithms. The evaluation confirmed that the SDK decreased execution time for ADAS algorithms and works seamlessly across different embedded architectures.

Aiming at the data sparse and cold start problems in collaborative filtering recommendation algorithm, an optimized solution based on user characteristics and user ratings is proposed in this paper. Based on users' basic attributes and users' history score record, the similarity of users and the similarity of items are calculated, and the nearest neighbor users and similar items are obtained. The advantage of the algorithm is that it combines the user's score and personal attributes to calculate the similarity between users and to recommend items. The optimized algorithm is applied to the recommendation of insurance products. Experiments based on real data from insurance company show that this method can reduce the average absolute error and improve the accuracy of recommendation.

According to the characteristics of Einstein wurfelt nicht! (EWN), studied the game's best and worst starting positions and some game strategies. In addition, this paper presents an effective optimized Monte Carlo algorithm based on the game strategies of EWN. Finally, the optimized Monte Carlo algorithm played chess respectively with the Monte Carlo algorithm, then a result got by comparative analysis.

Fast Oil Paint Image filter is a performance optimized oil paint algorithm. Current oil paint algorithms are CPU intensive and take a long time to produce the output; further the time taken increases exponentially with increasing quality. One of the main causes is re-computation. The proposed algorithm significantly reduces re-computation reducing the processing time by approximately 86%. The research also evaluates the algorithm using the parallel programming approach.

The sentiment information implied in the content of Social Networking Services (SNS) is of great technical and social significance, which has attracted a lot of researchers in different fields. Chinese text sentiment analysis is still in its starting stage, and most of the previous sentiment analysis algorithms are either inaccurate or inefficient. The purpose of this paper is to propose an algorithm performing well in both accuracy and efficiency, and apply it into a real time platform. A sentiment analysis algorithm on Chinese micro-blog content is introduced firstly, which achieves an outstanding accuracy but performs badly in efficiency. Then we optimize it with three strategies: data structure optimization, query strategy optimization, and parallel optimization. The experiment shows these strategies are very effective and the optimized algorithm is over 100 times more efficient than the basic algorithm. Based on the optimized algorithm, a text sentiment analysis platform is developed for real time sentiment. The platform offers two main functions including text sentiment analysis and user sentiment timeline. The results are outputted through a group of open APIs, which can be revoked by other developers and reduces their development costs. As a whole, the platform performs well on responding rate, capacity of concurrent users, stability and expandability.

In the public transport scheduling, the problem of transferring buses in stations becomes more and more important. Therefore, many people begin to research the number of the encountered buses. This work elaborates that how to ensure more encountered buses in a station, which can make it convenient to transfer buses. In this paper we build a model to optimize the buses transferring in one area. The model considers the changing of the time that one bus remains in a station, which can affect the encounter of two buses. And the model has a variable number which describes the resident time of one bus in a station and two objective functions that one is the number of the encountered buses and the other is the number of stations in which buses encounter. It calculates the maximum of the number of the encountered buses and the min of the number of encountered stations. In this way, it ensures the best departure time of all routes. So that it makes the cost and the time to transfer buses become less. In addition, it calculates the value by using the algorithm optimization. Through the experiment result, we can see that this algorithm can improve the efficiency of riding buses and the level of the services in the field of public transport.

The genetic algorithms (GAs), as the optimization methods, are expansively used in mobile robot techniques. In this paper a previously developed, multi-agent based map building process optimizing GA [1] will be revised in form of GA algorithm optimization and GA's parameter tuning. The results of the algorithm-improving will be compared and analyzed in the conclusion.

Weighted frequent pattern mining is suggested to discover more important frequent pattern by considering different weights of each item, and closed frequent pattern mining can reduces the number of frequent patterns and keep sufficient result information. In this paper, we propose an efficient algorithm DS_CWFP to mine closed weighted frequent pattern mining over data streams. We present an efficient algorithm based on sliding window and can discover closed weighted frequent pattern from the recent data. A new efficient DS_CWFP data structure is used to dynamically maintain the information of transactions and also maintain the closed weighted frequent patterns has been found in the current sliding window. Three optimization strategies are present. The detail of the algorithm DS_CWFP is also discussed. Experimental studies are performed to evaluate the good effectiveness of DS_CWFP.

The problem of massive data 3D-visualization is to use the algorithm optimization techniques in a certain hardware conditions, to keep a 3D visualization software run normally and maximize the speed of response of the software on the user's instruction in the process of dealing with massive data. The essence of the massive data 3D-visualization is rooted in that the computer hardware can not meet data processing requirements. The software technology in the algorithm can only make the software handle more data in the same hardware condition. There are two important problems in the massive data D-Visualization process which are storage capacity bottleneck and computing capacity bottlenecks. Massive data dynamic scheduling technology and LOD technology are the measures to react the above two problems.

In order to improve the traditional terrain simulation with the problem of high computation cost and poor visual effect, a novel eroded terrain simulation technique for computer games is developed. Water Map and sediment map are utilized for the improvement of traditional hydraulic erosion algorithm to achieve more convincing simulated effects. Moreover, four acceleration methods are employed for algorithm optimization to ensure real-time simulation. Experiment results show that the new technique is robust and efficient to simulate photorealistic eroded terrain for real-time computer games and animations.

In this paper, an algorithm optimization is made with improved simulated annealing algorithm aimed at data transmission strategy of streaming media System in P2P networks. According with global layout, we have established a mathematic model for data transmission strategy of streaming media System in P2P networks named VMDTSA-P2P, which completely have considered the affection of both the out-bandwidth and the available time. We also optimize the solutions with improved simulated annealing algorithm. Compared with existing data transmission strategies, this improved algorithm accelerates the speed of new nodes choose the data provider from its partner nodes when there are lots of users and insures the solution of the data provider is approximate the best. Thus, the play delay of the streaming media System is decreased and the quality of video playback is increased. By realizing it in a simulated P2P system, the accuracy and the efficiency are both verified in the simulation.

In this paper, we use an improved algorithm of artificial fish, and did the optimization in setting the border in the simulation platform, especially in the field of choosing ways of the robots, we use the Multi-threshold to reduce the uncontrollable actions when robots are in the game. And this method gives us an acceptable way to solve the issue.

This article addresses the Computers in Cardiology Challenge 2001 for predicting the onset of paroxysmal atrial fibrillation (PAF) from the surface electrocardiogram (ECG). To predict PAF, we developed an algorithm based upon the number and timing of atrial premature complexes (APCs) in the ECG. The algorithm detects classical isolated APCs, then predicts PAF based on a measurement of APC rate that favors the most recent APCs. The challenge database consists of 100 pairs of 30-minute ECG segments that may or may not directly precede a PAF episode. We used the learning set of the challenge database to optimize our algorithm. On the test set, it achieved scores of 40 out of 50 for PAF screening (event 1) and 44 out of 50 for PAF prediction (event 2).

This paper proposes a simple yet effective way of building a simulation tool for predicting the performance of disk drive servo systems for high TPI (tracks per inch) drives at the early stage of their development. The simulation tool is built based on the identification of the plant model and mechanical disturbances of existing drives. It is a powerful tool for servo algorithm optimization and selection of critical parameters such as the sampling rate.

Today there is a growing interest in the integration of health monitoring applications in portable devices necessitating the development of methods that improve the energy efficiency of such systems. In this paper, we present a systematic approach that enables energy-quality trade-offs in spectral analysis systems for bio-signals, which are useful in monitoring various health conditions as those associated with the heart-rate. To enable such trade-offs, the processed signals are expressed initially in a basis in which significant components that carry most of the relevant information can be easily distinguished from the parts that influence the output to a lesser extent. Such a classification allows the pruning of operations associated with the less significant signal components leading to power savings with minor quality loss since only less useful parts are pruned under the given requirements. To exploit the attributes of the modified spectral analysis system, thresholding rules are determined and adopted at design- and run-time, allowing the static or dynamic pruning of less-useful operations based on the accuracy and energy requirements. The proposed algorithm is implemented on a typical sensor node simulator and results show up-to 82% energy savings when static pruning is combined with voltage and frequency scaling, compared to the conventional algorithm in which such trade-offs were not available. In addition, experiments with numerous cardiac samples of various patients show that such energy savings come with a 4.9% average accuracy loss, which does not affect the system detection capability of sinus-arrhythmia which was used as a test case.