Search

Intra Block Copy (IBC) cannot work well for blocks with symmetry or mirror characteristic, because it is based on the translational block matching technique. In order to solve this problem, in this paper, an adequately designed IBC-Mirror mode is proposed for Versatile Video Coding (VVC), which extends our previous work. The newly proposed IBC-Mirror mode consists of two sub-modes, including IBC-Mirror Advanced Motion Vector Prediction (AMVP) mode and IBC-Mirror Merge mode. In IBC-Mirror AMVP mode, various Motion Estimation (ME) methods are applied, including previously proposed spatial search, and newly designed methods, i.e., Hash-based, Cache-based, Predictor-based and History-based IBC-Mirror search. As for IBC-Mirror Merge mode, Motion Information, composed of Block Vector (BV) and mirror flipping information, is added to the merge list to improve coding efficiency. Compared to the traditional IBC mode in VTM-10.0, the latest test model, the proposed mode can achieve 1~2% BD-rate saving. Compared to our previous work, methods in this paper can further achieve 0.5% coding gains and reduce 4% coding time. Higher coding efficiency and lower complexity show the superiority of extensions and improvements in this paper.

It is well-known that high-frequency information (e.g. textures, edges) is significant for single image super-resolution (SISR). However, Existing of deep Convolutional Neural Network (CNN) based methods directly model mapping function from low resolution (LR) to high resolution (HR), and they treat high-frequency and low-frequency information equally during feature extraction. Therefore, the high-frequency learning mode can not be sufficiently attentive, resulting in inaccurate representation of some local details. In this study, we aim to build potential frequencies’ relations and handle high-frequency and low-frequency information differentially. Specifically, we propose a novel Frequency Separation Network (FSN) for image super-resolution (SR). In FSN, a new Octave Convolution (OC) is adopted, which uses four operations to perform information update and frequency communication between high frequency and low frequency features. In addition, global and hierarchical feature fusion are employeed to learn elaborate and comprehensive feature representations, in order to further benefit the quality of final image reconstruction. Extensive experiments conducted on benchmark datasets demonstrate the state-of-the-art performance of our method.

Compare to rigid manipulators, flexible link manipulators (FLM) have the advantage of being lightweight and low energy consumption, which makes them have increasing demands across the industrial and aerospace world. Inverse dynamics-based control has been applied widely in rigid manipulators. However, the natural undamped behavior of FLM's internal dynamics hinders the direct use of inverse dynamics-based control. In this paper, an inverse dynamics-based control with parameter adaptation has been proposed, which is designed according to a redefined dynamic model with uncertainties for achieving good tip-tracking with good vibration attenuation as well. The proposed controller consists of model compensation based on inverse dynamics, feedback control, and internal control. The stability of the closed-loop system has been proved through Lyapunov theory. The experimental results indicate best performances in both slow and fast point-to-point motion with the proposed control.

In recent years, several optimal airline ticket purchasing strategies have been proposed and most of them are based on predicted price information. However, risks of ticket sold out and price uncertainties have not been modeled or considered by these strategies, which limits their effectiveness. Therefore, we design a recommendation service that is based on a new optimal airline ticket purchasing strategy PACES (Path Cost Expectation-based Strategy), which models and considers the risks and uncertainties of ticket purchasing explicitly. Extensive experiments are conducted on a real-world dataset and the results show that this strategy significantly outperforms the state-of-the-art techniques.

Coauthorship prediction is challenging yet important for academic collaboration and novel research topics discovery. The challenges lie in the dynamics of social or organizational relationships, changing preferences of suitable collaborators, and the evolution of research interests or topics. However, most current approaches and systems developed so far are mainly based on past coauthorships from a static viewpoint and do not capture the above evolving characteristics in coauthoring. Accordingly, this paper proposes a time-aware approach to capture the evolving coauthorships from online academic databases in terms of capturing the dynamics of social relationships and research interests. In particular, in order to understand the underlying factors influencing researchers to make choices of coauthors, we incorporate choice modeling based on utility theory. More specifically, our model conducts a series of pairwise choices over a poset induced by a utility function so as to learn the preference over all candidate coauthors. To complete the model inference, a gradient-based algorithm is devised to efficiently learn the model parameters for large-scale data. Finally, extensive experiments conducted on a real-world dataset show that our approach consistently outperforms other state-of-the-art methods.

© 2013 IEEE. In this paper, an Advanced Quantum-based Neural Network Classifier (AQNN) is proposed. The proposed AQNN is used to form an objectionable Web content filtering system (OWF). The aim is to design a neural network with a few numbers of hidden layer neurons with the optimal connection weights and the threshold of neurons. The proposed algorithm uses the concept of quantum computing and genetic concept to evolve connection weights and the threshold of neurons. Quantum computing uses qubit as a probabilistic representation which is the smallest unit of information in the quantum computing concept. In this algorithm, a threshold boundary parameter is also introduced to find the optimal value of the threshold of neurons. The proposed algorithm forms neural network architecture which is used to form an objectionable Web content filtering system which detects objectionable Web request by the user. To judge the performance of the proposed AQNN, a total of 2000 (1000 objectionable + 1000 non-objectionable) Website's contents have been used. The results of AQNN are also compared with QNN-F and well-known classifiers as backpropagation, support vector machine (SVM), multilayer perceptron, decision tree algorithm, and artificial neural network. The results show that the AQNN as classifier performs better than existing classifiers. The performance of the proposed objectionable Web content filtering system (OWF) is also compared with well-known objectionable Web filtering software and existing models. It is found that the proposed OWF performs better than existing solutions in terms of filtering objectionable content.

$k$ -nearest neighbors algorithm ( $k$ -NN) is a simple yet powerful method that can predict an unlabeled object using the $k$ closest samples whose labels are already known. However, in some cases such as medical treatment decision making, multiple experts can provide their individual decisions and there is no single ground truth for the labels of samples. As a result, the accuracy of the predictions can be affected by the uncertainties of historical decisions. In order to learn from uncertain decisions from multiple experts, we propose a new $k$ -NN algorithm. This algorithm firstly measures the uncertainties of labels. Then, the prediction is made based on the uncertain labels of the $k$ closest samples based on the Dempster-Shafer theory. We also introduce two methods to compute the optimal $k$ value. Experiments on a real-world medical treatment dataset prove that our algorithm has better performance compared with other algorithms.

While language-learning apps have empowered many users to learn languages inexpensively, the underlying technology limits the tasks they can offer to atomic, flashcard-like exercises. In this paper we present an algorithm that can effectively estimate memory traces from user interaction data that is generated in complex language learning tasks such as extensive reading. This technology can be applied to develop complex, intrinsically-motivating, high-value learning tasks based around long texts, television series or podcasts.

As water and land transportation hubs, container terminals are important infrastructures for national economic and social development and participation in global competition and cooperation. With the rapid economic development, the container throughput of China’s ports is increasing year by year. Accurate prediction of container terminal operation can optimize the terminal resource scheduling and improve the operation efficiency. Therefore, in this paper, we design a system to predict the container terminal operations. Some machine learning and deep learning algorithms are used to predict the number of vessels arriving at the port in the future days and the length of time container barges will be at port. Taking Shanghai Yangshan Port phase IV automated terminal as an example, we also perform some experiments, and the results show that our system can provide accurate predictions for the container terminal operations.

Content-based publish/subscribe systems provide subscribers with the ability to express their interests, but event matching operations need to be performed in a timely manner. To achieve a high matching speed, many efficient algorithms have been proposed. However, the performance of most algorithms is affected by the matching probability of subscriptions with events. Aiming to provide a fast and stable matching performance, in this paper, we propose an effective composite matching framework called Comat which explores the idea of performing event matching using multiple algorithms possessing complementary behavior to the matching probability of subscriptions. Comat predicts the matching time of the algorithms for each event and chooses the one that matches the event with the minimal cost. We implement a prototype of Comat based on two existing algorithms and TensorFlow. The experiment results show that Comat improves and stabilizes the matching performance by 25 % and 30 % respectively on average, well achieving our design goal.

In anti-air and anti-missile system, this paper aims at the severe threaten by anti-radiation weapon. It provides five ways against anti-radiation weapon, including pre-warning, cajolery, intercept to destroy, adopting advanced radar system and radar set net. Under protection of our army electronics equipment and personnel’s safety, we should strike electronics equipment and personnel of the enemy with all our might.

Recent years have observed the rapid development of astronomy observation devices, hence leveraging a large amount of observation data to automatically detect flare has become an emerging research topic. Previous studies on the flare detection task focus on using hand-drafted astronomy features or time-series analysis to capture the abnormal values in the luminosity data. However, these approaches heavily rely on domain expertise and are difficult to transfer into other stars or special phenomena. In this paper, we consider adopting deep learning technology into this task. To enhance the transferability and build an effective model, we propose a novel task, namely a masked difference prediction task to learn the enhanced representations of each luminosity difference and the whole sequence. The learned representations can be transferred into conventional RNN and CNN models with simply fine-tuning on the original flare detection task. Experiments show that our approach can bring improvement to CNN and RNN models.

This paper proposes an IBC-Mirror mode for Screen Content Coding (SCC) for the next generation video coding standards, including Versatile Video Coding (VVC) and Audio Video Standard-3 in China (AVS3). It is the first time to take mirror characteristic into consideration for SCC in VVC/AVS3. Based on the translational motion model of Intra Block Copy (IBC) mode, the function of "horizontal and vertical flipping" is further added to reduce prediction error and improve coding efficiency. The proposed IBC-Mirror mode is implemented on the latest reference software, including VTM5.0 (VVC) and HPM-5.0 (AVS3). The simulations show that the proposed mode can achieve up to 1~2% (VVC) and 4~7% (AVS3) BD-rate saving for SCC test sequences. Drafts about the mode have been submitted to AVS meeting and investigated in SCC Core Experiments (CE).

Based on the battlefield environment support information of sky and firmament, this paper puts focus on the study of anti-air missiles synergic battle effectiveness technique evaluation. By the analyses of function, usage of soft technique can avoid imitation of former static and dynamic effect valuation system, and make the analyzing results approach a true circumstance more. So, it has certain instruction meaning to the effectiveness improvement of anti-air and anti-missile battle.

This paper introduces several commonly used methods of target detection, using the difference between several adjacent frame images before and after the image sequence to extract the time difference method of the motion area in the image, using the difference between the current frame image and the background image to detect the background subtracting method of the motion area. After comparing the advantages and disadvantages of various methods, this section presents an algorithm that combines the combination of Gauss model algorithm and background subtract, which assumes that the observations of a pixel in an image are independent of other pixels, and then determines the foreground and background based on the probability of pixels.

As an informal language, text with dirty words is very common in online comments. Because of their adverse effects on communication users, text analysis with dirty words must be automated. In this paper, we first crawled past reviews that contained dirty words on many networks. Second, we manually marked 20,000 texts to form a corpus containing dirty words. Then use the data to train our model. We use the Decision tree classifier to detect dirty words contained in Malay text. Empirical research shows that our model works well. The F1-score of labels 0 and 1 reached 98% and 93%, and the precision rate of both is over 95%.

Deep neural network (DNN) is an important technology in the field of artificial intelligence. The softmax layer is one of the key component layers for completing multi-classifcation tasks. The softmax layer contains a large number of exponents and division operations, which causes it to consume a lot of hardware resources in FPGA implementation. This paper presents an efficient FPGA implementation of the softmax layer. For the exponential operation of the softmax layer, the mathematical transformation better than the Coordinate Rotation Digital Computer (CORDIC) algorithm is used to convert to subtraction and 2 exponential operation. The division is converted into a subtraction and shift operation, and added the highest non-zero bit detection processing before the operation to greatly reduce the cycle of the division operation. Experimenting with our softmax architecture on Xilinx ZCU102, the results show that the FPGA implementation of softmax layer can attain the precision of magnitude 10-6 and greatly accelerate the calculation of the softmax layer.

In recent years, with the development of Neural Network, it has made a significant breakthrough in the field of text detection and recognition. However, large-scale deep Neural Network needs a large amount of storage space and computing resources, which makes it difficult to be effectively applied in various hardware platforms. In order to solve this problem, we quantize the text detection and recognition model from floating-point to low bit-width. Utilizing the low bit-width quantization method in DoReFa-Net, we quantize the text detection network, Connectionist Text Proposal Network (CTPN) and the text recognition network, Convolutional Recurrent Neural Network(CRNN). Comparing the results in different quantized bit-width, we find that after quantizing the weights and activations of CTPN to 2-bit, the F-score only drops by 0.09 on the MTWI 2018 Dataset. When the weights and activations of CRNN is quantized to 4-bits, the accuracy decrease of training with Synthetic Chinese String Dataset is merely 0.1%. After the compression of the text detection and recognition, the size of model is reduced to only 32% of the original, which means that storage space can be saved by 68%.

In recent years, the organization of open source software development has evolved rapidly. Analyzing the behaviors of contributors helps us understand the development process of open source software and explore the general and special rules of it. Agent based model is a type of computing model that can simulate the behaviors and interactions between autonomous entities. With agent based models, we can simulate the self-organization process of open source software development by simulating contributors' behaviors. Therefore, we design an agent-based simulation system for open source software development, which are implemented with the Java Agent Development framework. In experiments, we obtain the simulation results by inputting historical behavior information of open source repositories on GitHub. Then we compare the results of various models. Finally, by adjusting input parameters for issue resolution process, we analyze the impact of these parameters, which shows this system also helps understand how to control the open source software development process.