Search

Inspired by the core idea of AlphaGo, we combine a neural network, which is trained by Adaptive Dynamic Programming (ADP), with Monte Carlo Tree Search (MCTS) algorithm for Gomoku. MCTS algorithm is based on Monte Carlo simulation method, which goes through lots of simulations and generates a game search tree. We rollout it and search the outcomes of the leaf nodes in the tree. As a result, we obtain the MCTS winning rate. The ADP and MCTS methods are used to estimate the winning rates respectively. We weight the two winning rates to select the action position with the maximum one. Experiment result shows that this method can effectively eliminate the neural network evaluation function's “short-sighted” defect. With our proposed method, the game's final prediction result is more accurate, and it outperforms the Gomoku with ADP algorithm.

In this paper a deep reinforcement learning (DRL) method is proposed to solve the control problem which takes raw image pixels as input states. A convolutional neural network (CNN) is used to approximate Q functions, termed as Q-CNN. A pretrained network, which is the result of a classification challenge on a vast set of natural images, initializes the parameters of Q-CNN. Such initialization assigns Q-CNN with the features of image representation, so it is more concentrated on the control tasks. The weights are tuned under the scheme of fitted Q iteration (FQI), which is an offline reinforcement learning method with the stable convergence property. To demonstrate the performance, a modified Food-Poison problem is simulated. The agent determines its movements based on its forward view. In the end the algorithm successfully learns a satisfied policy which has better performance than the results of previous researches.

This paper presents a visual attention based convolutional neural network (CNN) to solve the image classification problem in the real complex world scene. The presented method can simulate the process of recognizing objects and find the area of interest which is related with the task. Compared with the CNN method in image classification, the model is proficient in fine-grained classification problem and has a better robustness due to its mechanism of multi-glance and visual attention. We evaluate the model on vehicle dataset, where its performance exceeds CNN baseline on image classification.

The video surveillance system is widely used in various areas. However, it is an intractable problem to process a large number of video data manually. There are many intelligent systems developed to analyze human behavior and detect unusual events, however, few concerns the cheating behavior in examinations. In this paper, we use video surveillance system to detect cheating behavior in the examination. The pose estimation method is presented to detect the cheating behavior, based on the analysis of the implementation environment. More in detail, the pictorial structure is adopted to model the human body and a required pose and skin color information is utilized to build the human appearance model. Finally, the belief propagation algorithm is used to infer the maximum a posterior pose. The experiment shows the effectiveness of our method.

The march-fight transform equipment pull rod is the key component for some type anti-aircraft gun. If the pull rod is broken, the safety of operation staff will be seriously threatened, and normal combat readiness training will be affected. In addition, the firing stability and movement flexibility of the anti-aircraft gun will be affected. Firstly, based on the army survey, five hypothesis of failure cause for march-fight transform equipment pull rod were put for ward, including th at th e operation is unproper, that mate rial is ageing, that heat treatment is dissatisfied requirement, that the design is unreasonable an dother comprehensive f actor. Secondly, th e eleven pull rods were extracted from difference factory, and the broken on es were p ut u p mechanics analyse. Ac cording to th e result, breaking cause of th e march-fight transform equipment pull rod for t he an ti-aircraft gun were analysed. Fin ally, the following conclusion were reached. On e is that the pull rod design exists the defect; Second is that some pull rod mate rial and heat treatment falls short of the requirement. And according to the conclusion, the design modification and improvement advise are put forward such as the standard operating rules, heat treatment, thermal refining and the improvement design of pull rod etc. This research greatly enahanced the reliability and safety of the march-fight transform equipment pull rod for some type anti-aircraft gun.

Recently, the Ad Hoc networks have increased their popularity and kept the interest of the most important research groups all over the world due to their automatically and quickly self-organization. But the restricted energy of mobile hosts is the bottleneck of the network lift time. Based on the hierarchical Ad Hoc network, an adaptive mechanism for building network based on energy is proposed. This mechanism considers the influence of the energy of the whole network and mobile hosts adequately; it implements dynamically choosing cluster heads, balancing the burden of cluster heads, and also can improve the connectivity of links. The mathematical demonstration, simulation results based on the change of the total number of mobile hosts and the initial degrees, and the comparison with the algorithm of most link degrees show that this proposed mechanism possesses better degree distribution of cluster heads and can assign the energy of the whole network reasonably and prolong the lifetime of the network, and also improve the performance and communication efficiency partially.