Toshimichi Saito

Design of switching circuits based on particle swarm optimizer and hybrid fitness function

This paper studies application of the particle swarm optimizer to finding suitable parameters for multi-objective problems in design of switching circuits. The problem is described by the hybrid fitness function consisting of analog objective functions of the parameters, criterion and digital logic. The hybrid fitness function permits increase of some fitness component(s) below the criterion and this flexibility can realize good performance. We then consider an example: optimization of switching angles of the inverters. In this application, each particle corresponds to switching angles of the inverters and moves to improve the total harmonic distortion and average power. Performing basic numerical experiments, the algorithm efficiency is confirmed.

PSO-based multiple optima search systems with switched topology

This paper discusses a particle swarm optimization (PSO) with switched topology and its application to the multi-solution problems. First, we introduce a deterministic PSO characterized by normalized deterministic parameters and a canonical form system equation. This system is convenient to grasp effects of parameters on the stability. Second, we investigate effects of the average distance of several the swarm topologies on the search capability. Especially, we introduce the switched topology where any information is not transmitted from the edge if the switch is off. Third, we consider an application to exploring multiple periodic points in simple dynamical systems. Performing numerical experiments for typical examples, the algorithm performance is investigated.

Application of particle swarm optimizers to two-objective problems in design of switching inverters

This paper presents an application of the particle swarm optimizers to multi-objective problems in design of switching inverters. In the algorithm, the particles represent switching angles and move to optimize two objectives: minimizing the total harmonic distortion and setting the desired total power. This two-objective problem is important to realize smooth speed control of ac machine drivers. Performing basic numerical experiments, we have confirmed that the method can find efficient switching angles that realize efficient operation.

A GA-based flexible learning algorithm with error tolerance for digital binary neural networks

This paper presents a learning algorithm of digital binary neural networks for approximation of desired Boolean functions. In the learning, the genetic algorithms is used with flexible fitness that tolerates error: it is suitable to reduce the number of hidden neurons and to tolerate noise and outliers. We then apply the algorithm to design of cellular automata with rich spatio-temporal patterns and various applications. Performing basic numerical experiment, the algorithm efficiency is confirmed.

Learning of dynamic BNN toward storing-and-stabilizing periodic patterns

This paper studies learning algorithm of a dynamic binary neural network having rich dynamics. The algorithm is based on the genetic algorithm with an effective kernel chromosome and hidden neuron sharing. Performing basic numerical experiments, we have confirmed that the algorithm can store desired periodic teacher signals and the stored signals are stable for initial value.

Dynamic binary neural networks and evolutionary learning

This paper studies the dynamic binary neural network having N bits input, N bits output and ternary weighting parameters of the hidden layer. Applying feedback from the output to the input, the network can generate dynamic binary sequence. We presents a simple learning algorithm that uses the genetic algorithm and reduces the number of hidden neurons efficiently. Performing a basic numerical experiment, the algorithm efficiency is confirmed. Application to switching power converters is also discussed.

Parallel ant colony optimizers with local and global ants

This paper studies the ant colony optimizer with parallel processing function based on adaptive resonance theory map. The optimizer has two groups of ants: local ants that is assigned to search in a subspace and global ants for global search. Effective communication between local and global ants is key to realize desired optimization. Applying the algorithm to typical bench marks, we can suggest that the optimizer can realize adaptive and fast search of solutions.

Growing Particle Swarm Optimizers with a Population-Dependent Parameter

This paper studies a new version of growing particle swarm optimizers. In the algorithm, a new particle is born if a particle exploring the optimum is stagnated and the swarm can grow depending on problem complexity. The particle velocity is controlled by an acceleration parameter that can attenuate depending on the number of particles and can vibrate depending on the time. The parameter plays important role to reduce the computation cost and to increase the success rate. The algorithm efficiency is confirmed by numerical experiments of typical benchmarks.

Analysis of stability and bifurcation in a simple model of power converters with solar cell input

This paper studies stability and bifurcation of a simple piecewise linear dynamical system relating to several kinds of switching converters with solar cell input. Applying a simplification method, we derive the 1D phase map that is useful for precise analysis. The system can exhibit rich bifurcation phenomena, and for simplicity, we consider bifurcation for two parameters corresponding to temperature and an inner conductor. We then provide two important results. First, as the temperature parameter increases, stable periodic orbit is changed into chaotic orbit via period doubling bifurcation. This bifurcation set is a border between stable operation and chaos generation. Second, as the conductor parameter approaches zero, the chaotic orbit is changed into a variety of super-stable periodic orbits.

Delay-induced order in pulse-coupled bifurcating neurons

This paper studies dynamics of pulse-coupled bifurcating neurons with delay. Before the coupling, the neuron can exhibit chaotic/periodic behavior by repeating integrate-and-fire behavior between the threshold and sinusoidal base signal. After the coupling, the system can exhibit various bifurcation phenomena. Especially, we have found an interesting phenomenon: chaotic behavior of single neuron can be changed into periodic behavior of coupled system by a delay effect. This delay-induced order and related bifurcation can be analyzed precisely using the mapping procedure. Presenting a simple equivalent circuit, basic phenomena are confirmed experimentally.