Search

To solve the memory losses of raw information due to the deepening network layers, a model based on memory correction mechanism is adopted to address the attention bias and the raw information losses. By introducing memory correction mechanism to GA Reader, option coding module is applied to RACE dataset for solving the issue that datasets have fewer demands for deep reasoning ability and contain noises. According to the experiment and analysis, the optimized model has each layer fused with the raw document information in the process of network iterative learning to keep correcting attention for the iterating network. Meanwhile, the reasoning ability needed for the model has also been improved. The RACE machine selection reading comprehension model based on memory correction could effectively solve the issue of raw information losses and overly simplified datasets.

Recommender systems have proved to be an important response to the information overload problem, by providing users with more proactive and personalized information services. Collaborative filtering is the most popular method in implementing a recommender system. The Slope One algorithm, which is one of collaborative filtering algorithms, is not only easy to implement, but also efficient and effective. However, the prediction precision of the Slope One algorithm is low. It is important to improve the prediction precision when data is sparse. This paper introduces a trustbased recommend model upon collaborative filtering. Firstly, the fraud users among the raw data set are spotted by employing Bayesian method, and then those fraud ratings are removed; secondly, the degree of trust between users and the degree of trust for ratings itself are calculated; finally, the trust-based recommend model upon collaborative filtering is proposed. We can conclude that the improved Slope One algorithm by employing the trust-based recommend model has a more accurate prediction precision than the traditional Slope One algorithm which is based on user similarity.

A novel deployable array mechanism is proposed in this paper which provides a new concept for the future large SAR antenna for a spacecraft. This deployable array mechanism, which is composed by several main array panels and associate array panels, is able to expand in two dimensions with only one degree of freedom. The structure and the skeleton of the novel deployable mechanism are described at first. And then, the topology and the mobility of this mechanism are presented and investigated. To avoid the interference, the geometric size of the associate array panels has been studied. For the purpose of obtaining the driving torque conveniently, the simplified dynamics model for this mechanism has been established. Finally, the simulations are conducted to validate the feasibility and correction of the above model and analysis. And the improvement for this structure inspired by origami art is presented. The achievement in this research helps to develop the new deployable mechanisms in aerospace field.

In order to overcome the co-channel interference between neighbor cells in Long Term Evolution (LTE) system, a novel power allocation algorithm based on non-cooperative game theory is proposed in this paper. In the downlink of LTE, the pricing mechanism in the non-cooperative game theory is used to model the power allocation scheme in Orthogonal Frequency Division Multiple Access (OFDMA) cellular systems. We set the data rate as the pricing factor, and the objective function is the variance of power. Then we use Genetic algorithm to obtain the optimal solution. Both the QoS of user and the system throughput are considered in our proposed algorithm, simulation results show that the co-channel interference is reduced maximally. Comparing with the traditional water filling algorithm, the user can obtain more system throughput as the number of user increasing. It is suitable for the crowded user areas.

By the ways and theory on agriculture ece-economic system, in the paper, Gonglian village agriculture ece-economic systems of Hengnan county in 2002 and 2004 were selected as examples, author counted the agricultural material input and output taking N, P, K as the index to meas ure the material flows in agriculture ece-economic system, and toke analysis on function into the degree of material devotion, scale structure of devotion and the degree of input, etc.The results showed that it was high wholely that the function benefit of agriculture ece-economic system in Gonglian village, where the ecological agriculture optimization model is worthy to be promoted.

Automatic construction of a full view panorama is an active area of research in the fields of photogrammetry, computer vision and computer graphics. Using fisheye camera for panorama generation could greatly reduce the number of images thanks to its wide field of view. A key step in creating panorama with fisheye camera is to estimate the cameras' motion parameters which are usually estimated by established correspondences. However, establishing correspondences is a difficult thing particularly for high distorted images. For this reason, most of the traditional systems based on fisheye cameras for creating panorama are only designed for some specific configurations such that correspondences can be easily established. In this paper, we try to give a more general solution to establishing correspondences for a fisheye image pair. First, the feature detector unaffected by the non-linear distortion (like the MSER-regions) is used to extract features in original image. Then feature matching is realized on the virtual camera plane (VCP). Next the interested region is back-projected onto the VCP where it becomes a perspective one, so the standard affine invariant based matching algorithms can be employed directly. To combat data noise and improve the precision of the estimates of motion parameters, more correspondences are searched by applying guided matching strategy. Experiments showed that our algorithm can lead to the precise and reliable estimates and the final full view panorama is satisfactory.

Based on the Lagrangian equations of first type, the coordinate system is constructed according to the mechanical characteristics of Hexa Parallel Robot. The generalized coordinates are composed of six actuated independent coordinates plus redundant coordinates denoting the position and three Euler angles of active platform. After formulating the Lagrangian function, combining the constraint equations, the dynamics model of Hexa Parallel Robot is established.

Compared with conventional perspective camera, data noise for omnidirectional camera is more serious due to its non-linear projection. For linear estimation of motion parameters with noise corrupted data, it is necessary to establish sufficient number of correspondences and use a appropriate robust tech- nique. In this paper, our interests concentrate on two parts. First, we propose a new method to establish a large number of reliable point correspondences between omnidirectional images. Classical affine invariant region detecting and matching algorithms are initially employed, then more distinctive point correspondences inside or near the matched regions are extended. Our second work focuses on the RANSAC based linear estimation technique. Based on the simulation experiment we give a strategy to grad- ually select more and more inliers from a noise-corrupted data set, by which the initial guess of motion parameters improved in precision and stability. Experiment results with true data show that our method works well in real applications and makes the performance of linear estimates of motion parameters improved greatly.