فهرست مطالب سعید شیری قیداری

Semi-supervised metric learning has attracted increasing interest in recent years. They enforce smoothness label assumption on the manifold. However, they suffer from two challenges: (1) since data in each class lies on one manifold and the similarity between classes leads the intersection between manifolds, the smoothness assumption on the manifold is violated in intersecting regions. (2) 1NN classifier, which is applied for predicting the label of classes in metric learning methods, is suffered from the rare of labeled data and has not suitable accuracy. In this paper, a novel method for learning semi-supervised metric in the stratified space has been proposed that exploit the prior knowledge, which is the smoothness assumption on each manifold, more accurate than existing methods. In the metric learning stage, it doesn’t apply smoothness assumption on the intersecting regions and in the classification stage, labeled data in the interior regions of manifolds are extended based on the smoothness assumption. The different behavior of the Laplacian of piecewise smooth function on stratified space is exploited for the distinction of the intersecting regions from interior regions of manifolds. The results of experiments verify the improvement of the classification accuracy of the proposed method in the comparison with other methods.

This study deals with optimal operation of multipurpose reservoir systems, functioning for water supply to consumptive agricultural demands, hydropower generation and minimum environmental flow purposes, using cellular automata (CA). Discrete time instances (beginning of monthly time periods) are taken as cells and storage volumes at these instances are considered as the states of the cells. The CA updating rule is derived by optimization of local objective functions analytically in presence of nonlinear hydropower equations. Additionally, the sub-problem of optimal water allocation among different downstream users is solved using Lagrange multipliers technique. The proposed method was used to solve the nonlinear, non-convex operation optimization of Balarud reservoir and its hydropower plant as a real case study, and its performance was compared with the ones of gradient-based optimization methods and Genetic algorithm for periods of 5, 20 and 44 years. The results showed that the proposed cellular automata-based optimization model was able to find solutions with acceptable quality from the optimality perspective while rendering considerable reduction in the required computational load.

The electrohydraulic valves are commonly used in the engineering applications. These valves, as the medium elements, prepare the hydraulic systems for the electrical control applications. For the precise performance of these valves, disturbances in the valve elements dynamics will disturb the control process of the system. The electrohydraulic servo valves are greatly affected by the external acceleration, for instance in the aerospace applications. In a two stage flapper- nozzle electrohydraulic valve, the external acceleration changes the pressure of the fluid leaving the nozzles and it affects the flapper and spool of the valve like a virtual force. Thus, when the applied current is zero, the acceleration diverts the spool of the valve from the equilibrium point, and unwanted performance in the valve occurs. In this study the pilot pressures of the spool is modeled in unsteady state condition. The effects of the acceleration on the flapper and the spool of the two stage electrohydraulic valve are investigated. At the end, the obtained model is verified by use of the experimental data.

Beautiful and Meaningful Iranian Names Production by Genetic Algorithm using Artificial Neural Network-Based Fitness Function.

In this paper, the dynamic Bayesian network for modeling behavior has been used for door-crossing behavior. This network easily trains and it supports dependencies between of data’s sensors and the desired behavior of robot on ideal environment and it detects error on data’ sensor by using data based method in real environment. Elimination of undesired sensor performs with structure changing base on interpretation of sensor’s data. Evaluation is based on robots with artificial error in the readings of the sensors. Proposed model for the "door-crossing", which is based on dynamic Bayesian networks, has been able to resist much higher than the model based on Bayesian networks (fixed structure), in the face of adverse and multiple interpretations of the constant error, drift, also cross-talk, random movement and slight changes in the environment map. This model is much resistant by adding maneuver part to it on dynamic environments. The performance of the proposed model on the simulation environment has been evaluated.

In this paper a new structure based on Bayesian networks is presented to improve mobile robot behavior, in which there exist faulty robot sensors. If a robot likes to follow certain behavior in the environment to reach its goal, it must be capable of making inference and mapping based on prior knowledge and also should be capable of understanding its reactions on the environment over time. Old learning models for knowledge learning, especially on dynamic environment, are quite complex and have uncertainty in sensors. In this paper a new structure based on Bayesian network is presented for knowledge learning on robot behavior when the malfunction sensors exist. In this paper successful door crossing behavior is explained. In this issue the belt of ultrasonic sensors is used to receive environment information. Simulation results show that using the Bayesian network is very effective in robot behavior with faulty sensors.