Abstract: Cyclical activities are basic characteristics of all living organisms. Neurobiologists have discovered that a single neuron often possesses membrane properties that are responsible for the generation of oscillations. When coupled with other neurons, oscillations with varying properties depending on the type of interconnection can be generated. Using synchronization and temporal correlation of these oscillations can carry out the tasks of pattern recognition of different objects. The speed of recognition depends on the speed of synchronization. In this paper, we propose evolutionary coupled neural oscillators to minimize the time of synchronization through the optimization of the neuron parameters by means of a genetic algorithm. The genetic algorithm, with its global search capability, finds the optimum neuron parameters through a fitness measure that reflects the correlation strength between oscillators, thus avoiding the trial-and-error process of estimating the neuron parameters. The superiority of the method is demonstrated through an application of character recognition process.
Abstract: In many applications, the robot’s environment is changing with time in a way that is not predictable by the designer in advance. In addition, the information available about the environment is subjected to imprecision and incompleteness due to the limited perceptual quality of the sensors. These problems can be handled by combining the adaptive power of both evolutionary strategy algorithms and fuzzy controllers. In this paper, an evolutionary strategy algorithm is used to tune fuzzy membership functions to enhance the performance of a fuzzy controller that governs a robot behavior. This fuzzy controller is synthesized from human heuristics with respect to various situations of the changing environment. The controller acts according to a combination of both goal seeking and open area seeking approaches. The proposed system was evaluated through different simulations of the robot’s environment and it achieved promising results.
Abstract: An educational timetabling is a multi-dimensional and highly constrained problem. Generating educational timetables manually often involves numerous rounds of changes before they can be satisfactory. Usually such a process takes several days, and often the quality of the timetables is compromised due to pressure to release the timetables on time. Automatic generation of timetables then seems to be an attractive to manual approach. But this approach is not without problems. In fact, most timetabling problems are NP-complete and most researchers are interested in investigating efficient algorithms for solving the problem. In this paper, a university timetable problem formulation is introduced followed by recent approaches for solving the problem. After that, a genetic algorithm (GA) is presented to efficiently and effectively solve the problem. The proposed GA has a flexible representation that handles all the college timetables at once. It incorporates repair strategies to always guarantee the creation of a feasible timetable which satisfies constraints that must not be broken. The algorithm is implemented and applied to create timetables for the college of computing at the Arab Academy for Science and Technology (AAST) in Egypt and it shows promising results.