Zheng Tang

Four-quadrant CMOS current-mode multiplier independent of device parameters

In this work, we present a four-quadrant CMOS current-mode multiplier based on the square-law characteristics of an MOS transistor operated in the saturation region. One advantage of this multiplier is that the output current is independent of MOS transistor device parameters; another, that the input resistance is independent of the input current. Simulations of the multiplier demonstrate a linearity error of 1.22%, a THD of 1.54%, a -3 dB bandwidth of 22.4 MHz, and a maximum power consumption of 0.93 mW. Operation of the multiplier was also confirmed through an experiment using CMOS 4007 ICs.

An improved backpropagation algorithm to avoid the local minima problem

Abstract We propose an improved backpropagation algorithm intended to avoid the local minima problem caused by neuron saturation in the hidden layer. Each training pattern has its own activation functions of neurons in the hidden layer. When the network outputs have not got their desired signals, the activation functions are adapted so as to prevent neurons in the hidden layer from saturating. Simulations on some benchmark problems have been performed to demonstrate the validity of the proposed method.

Batch type local search-based adaptive neuro-fuzzy inference system (ANFIS) with self-feedbacks for time-series prediction

This paper presents an improved adaptive neuro-fuzzy inference system (ANFIS) for the application of time-series prediction. Because ANFIS is based on a feedforward network structure, it is limited to static problem and cannot effectively cope with dynamic properties such as the time-series data. To overcome this problem, an improved version of ANFIS is proposed by introducing self-feedback connections that model the temporal dependence. A batch type local search is suggested to train the proposed system. The effectiveness of the presented system is tested by using three benchmark time-series examples and comparison with the various models in time-series prediction is also shown. The results obtained from the simulation show an improved performance.

A modified error function for the backpropagation algorithm

Abstract We have noted that the local minima problem in the backpropagation algorithm is usually caused by update disharmony between weights connected to the hidden layer and the output layer. To solve this problem, we propose a modified error function. It can harmonize the update of weights connected to the hidden layer and those connected to the output layer by adding one term to the conventional error function. It can thus avoid the local minima problem caused by such disharmony. Simulations on a benchmark problem and a real classification task have been performed to test the validity of the modified error function.

Gravitational search algorithm combined with chaos for unconstrained numerical optimization

Gravitational search algorithm (GSA) is the one of the newest developed nature-inspired heuristics for optimization problem. It is designed based on the Newtonian gravity and has shown excellent search abilities when applying it to optimization problems. Nevertheless, GSA still has some disadvantages such as slow convergence speed and local optima trapping problems. To alleviate these inherent drawbacks and enhance the performance of GSA, chaos, which is of ergodicity and stochasticity, is incorporated into GSA by two kinds of methods. One method uses chaos to generate chaotic sequences to substitute random sequences, while another one uses chaos to act as a local search approach. The resultant hybrid algorithms, called chaotic gravitation search algorithms (CGSA1 and CGSA2), thus reasonably have advantages of both GSA and chaos. Eight widely used benchmark numerical optimization problems are chosen from the literature as the test suit. Experimental results demonstrate that both CGSA1 and CGSA2 perform better than GSA and other five chaotic particle swarm optimization.

An Improved Clonal Selection Algorithm and Its Application to Traveling Salesman Problems

In order to improve some fundamental problems of the clonal selection algorithm (CSA), a novel clonal selection algorithm (NCSA) is proposed. After analyzing the mechanism of the clonal selection and proposing the antibody model, the basic character of the application problem fused into the NCSA based on rearrangements of antibody molecule coding genes. Next, we analyzed synthetically the antibody-antigen affinity and the antibody-antibody affinity. Through contrast experiments between the NCSA and the CSA to solve travelling salesman problem (TSP), it is shown that the NCSA has better performances.

Ant colony optimization with clustering for solving the dynamic location routing problem

Ant colony algorithm can resolve dynamic optimization problems due to its robustness and adaptation. The aim of such algorithms in dynamic environments is no longer to find an optimal solution but to trail it over time. In this paper, a clustering ant colony algorithm (KACO) with three immigrant schemes is proposed to address the dynamic location routing problem (DLRP). The DLRP is divided into two parts constituted by a location allocation problem (LAP) and a vehicles routing problem (VRP) in dynamic environments. To deal with the LAP, a K-means clustering algorithm is used to tackle the location of depots and surrounding cities in each class. Then the ant colony algorithm is utilized to handle the VRP in dynamic environments consisting of random and cyclic traffic factors. Experimental results based on different scales of DLRP instances demonstrate that the clustering algorithm can significantly improve the performance of KACO in terms of the qualities and robustness of solutions. The ultimate analyses of time complexity of all the heuristic algorithms illustrate the efficiency of KACO with immigrants, suggesting that the proposed algorithm may lead to a new technique for tracking the environmental changes by utilizing its clustering and evolutionary characteristics.

Letter: A binary Hopfield neural network with hysteresis for large crossbar packet-switches

In this paper, we propose a hysteretic Hopfield neural network architecture for efficiently solving crossbar switch problems. A binary Hopfield neural network architecture with hysteresis binary neurons and its collective computational properties are studied. The network architecture is applied to a crossbar switch problem and results of computer simulations are presented and used to illustrate the computation power of the network architecture.

A learning method in Hopfield neural network for combinatorial optimization problem

Abstract In this letter, we utilize the Hopfield network learning method to adjust the balance between constraint term and cost term of the energy function so that the local minimum that the network once falls into vanishes and the network can continue updating in a gradient descent direction of energy. We applied the proposed learning method to the traveling salesman problem to show that the method is capable of finding an optimal solution or a near-optimal solution in a shorter time.

Multiple-valued immune network model and its simulations

This paper describes a new model of multiple-valued immune network based on biological immune response network. The model of multiple-valued immune network is formulated based on the analogy with the interaction between B cells and T cells in immune system. The model has a property that resembles immune response quite well. The immunity of the network is simulated and makes several experimentally testable predictions. Simulation results are given to a letter recognition application of the network and compared with binary ones. The simulations show that, beside the advantages of less categories, improved memory pattern and good memory capacity, the multiple-valued immune network produces a stronger noise immunity than binary one.