1School of math and statistics, Nanjing University of Information Science & Technology, Nanjing 210044, China (Received October 01 2017, accepted January 15 2018) Abstract 。 Accurate segmentation for brain magnetic resonance (MR) images is of great significance to quantitative analysis of brain image. However, traditional segmentation methods suffer from the problems existing in brain images such as noise, weak edges and intensity inhomogeneity (also named as bias field). Convolutional neural networks based methods have been used to segment images; however, it is still hard to find accurate results for brain MR images. In order to obtain accurate segmentation results, a multi-scale fully convolution networks model (MSFCN) is proposed in this paper. First, we use padding convolutions in conv-layer to preserve the resolution of feature maps. So we can obtain segmentation results with the same resolution as inputs. Then, different sized filters are utilized in the same conv-layer, after that, the outputs of these filters are concatenated together and fed to the next layer, which makes the model learn features from different scales. Both experimental results and statistic results show that the proposed model can obtain more accurate results.

(Received January 02 2018, accepted March 18 2018) Laser spot emitted by laser pointers were widely used in numerous fields such as pointing during presentations, guiding robots in human machine interaction (HCI), and guiding munitions to targets in military, etc. Our interest is to develop an effective method for laser spot detection and tracking with exploiting fewer features. In this paper, we jointly combined pyramid Lucas–Kanade (PLK) optical flow for detection and extended Kalman filter for accurately tracking laser spot in low-resolution and varying background video frames. Using this new technique, we only need to deploy intensity of laser spot with displacement of laser spot and get a very good result. This technique could be embedded into multi-core processing devices with inexpensive cost for potential high benefits in the future.

(Received January 18 2018, accepted March 11 2018) This paper presents a 1-2 suppliers-retailers model with delayed bounded rationality. The retailers have their own products, and their products aren’t manufacturer’s products’ substitutes, products are epiphytic relation- ship. The phenomenon of chaos and other complex phenomena are reported using stability region,bifu1rca- tion,attractors etc. We also introduce delayed decision into the model, study the influence of delayed decision on the stability of the model. The results show that the system’s stability is mainly determined by the delay coefficient, ap- propriate delay coefficient can enhance the stability of the system, the inappropriate delay coefficient will reduce the stability of the system. In addition, we cannot simply think that more merchants adopt delayed decision can improve the stability of the system.

Due to the existence of inertia and response time, time-delay always exists in the response of the guide vane opening and it may bring about multi time-scale effect. Therefore, time-delay modeling of hydro-turbine gov- erning system with multi time-scale is addressed to investigate the dynamical behaviors of hydro-turbine governing system. The effects of the multi time-scale and time-delay on the dynamical behavior of the proposed system are analyzed. The fast-slow characteristic of the system is discussed via numerical simulations with the change of time- scale or time-delay. Results suggest that not only the time-scale but also the time-delay have significant impact on the dynamical behaviors of the hydro-turbine governing system.

In this paper, based on the classic BA scale-free network model, we proposed a new evolution model that gives a more realistic description of the people’s behavior on social networks. In the process of growth, there are local preferential attachment mechanisms and random attachment or removal between the old and new edges. We proved that the extended model follows the power-law distribution and the power exponent is between 2 and 3, which provides a theoretical support for analyzing the similar social network. Compared with the classic BA model, the extended model has a smaller average path length and a larger clustering coefficient, which is more consistent with the real social network.

Employees’ performance appraisal is often considered as a core management task in most standard organizations because it could potentially inspire an employee towards the pursuit of organizational goals and objectives. Meanwhile, issues such as delay in appraisal processes, inevitable human error, and emotional status of the appraiser, are common with the traditional appraisal methods. Hence, this research proposes a fuzzy decision support system based on Henri Fayol’s 14 principles of management for the appraisal of employees’ performance. A questionnaire consisting of 44 structured questions was designed based on these principles and administered to the employees of Babcock University, Ilisan-Remo, Ogun State, Nigeria. Fuzzy inference system that incorporates Mamdani computational technique was built based on the feedback extracted from the questionnaires, and used for appraising the employee performance. Experimental results show that the proposed system could predict the appraisal status of an employee with 75% accuracy in comparison to the conventional appraisal method.

The immunization strategy towards the epidemic spreading problem has attracted widespread attention of scientists from many different fields. However, the traditional immune behavior is achieved by deleting the edges in the network, which can lead to variations in the network structure and furthermore, seriously damage the efficien- cy of the network. In this paper, we studied a new type of immune strategy applied to weighted networks– immunization therapy strategy, which is to maintain the necessary network efficiency by lowering the weight of edges to suppress the spread of epidemic. It is similar to the inflammation around the infected parts of our body, which can either prevent epidemic from further spreading or do no harm to the function of the body. Thus, we name this novel strategy as "immunization therapy strategy". We first let the rate of transmission be proportional to the edge weight according to the ? − ? epidemic spreading model. In addition, we give specific dynamic evolution for infected nodes that boosts efficient epidemic control. Theoretical analysis and simulation outputs indicate that the immunization therapy strategy can efficaciously prevent the spread of the epidemic, while maintaining the high effi- ciency of the network.

(Received March 06 2017, accepted February 11 2018) Abstract Use of internet increases day by day so securing network and data is big issue. So, it is very important to maintain security to ensure safe and trusted communication of information between different organizations. Because of these IDS is very useful component of computer and network security. IDS system is used by many organizations or industries to detect the weakness in their security, documenting previous attacks and threats and preventing all of this from violating security policies .Because of these advantages this system is important in system security. In this paper we find the solution for different approaches (attacks) based on intrusion detection system. In this paper we identify different attacks and find solution for different type of attacks such as DDOS, SQL injection and brute force attack. In this case we use client server architecture. To implement this we maintain profile of user and base on this we find normal user or attacker ,when system find that attack is present then it directly block the attack.

The physical meaning of the Ikeda model is analyzed. The influence of the parameters in the model on the state of the solution is discussed. The change process of the solution when the parameters are taken from different values is described, and the sensitivity of the solution to the initial value is studied.The tail filter is used to optimize the decision function through the gradient descent method, filtering the observation error, and obtaining the initial field which is closer to the true value.