A learning algorithm for radial basis function support vector machines (RBF-SVMs) that can be easily implemented in digital VLSI is proposed. It is shown that, as opposed to traditional artificial neural networks, lea...
详细信息
A learning algorithm for radial basis function support vector machines (RBF-SVMs) that can be easily implemented in digital VLSI is proposed. It is shown that, as opposed to traditional artificial neural networks, learning in SVMs is very robust with respect to quantisation effects deriving from the finite precision of computations.
Gelenbe has modeled neural networks using an analogy with queuing theory. This model (called Random Neural Network) calculates the probability of activation of the neurons in the network. Recently, Fourneau and Gelenb...
详细信息
Gelenbe has modeled neural networks using an analogy with queuing theory. This model (called Random Neural Network) calculates the probability of activation of the neurons in the network. Recently, Fourneau and Gelenbe have proposed an extension of this model, called multiple classes random neural network model. The purpose of this paper is to describe the use of the multiple classes random neural network model to learn patterns having different colors. We propose a learning algorithm for the recognition of color patterns based upon non-linear equations of the multiple classes random neural network model using gradient descent of a quadratic error function. Ttl addition, we propose a progressive retrieval process with adaptive threshold values. The experimental evaluation shows that the learning algorithm provides good results.
We present a new evolutionary algorithm-"learning algorithm" for multimodal optimization. The scheme for reproducing a new generation is very simple. Control parameters, of the length of the list of historic...
详细信息
We present a new evolutionary algorithm-"learning algorithm" for multimodal optimization. The scheme for reproducing a new generation is very simple. Control parameters, of the length of the list of historical best solutions and the "learning probability" of the current solutions being moved towards the current best solutions and towards the historical ones, are used to assign different search intensities to different parts of the feasible area and to direct the updating of the current solutions. Results of numerical tests on minimization of the 2D Schaffer function, the 2D Shubert function and the 10D Ackley function show that this algorithm is effective and efficient in finding multiple global solutions of multimodal optimization problems. (c) 2008 Elsevier Ltd. All rights reserved.
A new learning algorithm for pattern classification using cellular neural networks is described. The authors show that patterns belonging to the training set as well as patterns outside it can be classified reliably u...
详细信息
A new learning algorithm for pattern classification using cellular neural networks is described. The authors show that patterns belonging to the training set as well as patterns outside it can be classified reliably using the proposed algorithm. Comparisons with well established classification techniques clearly highlight the performances of the approach developed herein.
An algorithm for the training of multilayered neural networks solely based on linear algebraic methods is presented. Its convergence speed up to a certain limit of learning accuracy is orders of magnitude better than ...
详细信息
An algorithm for the training of multilayered neural networks solely based on linear algebraic methods is presented. Its convergence speed up to a certain limit of learning accuracy is orders of magnitude better than that of the classical back propagation. Furthermore, its learning aptitude increases with the number of internal nodes in the network (contrary to backprop). Especially if the network includes a hidden layer with more nodes than the number of examples to be learned and if the number of nodes in succeeding layers decreases monotonically, the presented algorithm in general finds an exact solution.
In this paper, first we propose an architecture of fuzzy neural networks with triangular fuzzy weights. The proposed fuzzy neural network can handle fuzzy input vectors as well as real input vectors. In both cases, ou...
详细信息
In this paper, first we propose an architecture of fuzzy neural networks with triangular fuzzy weights. The proposed fuzzy neural network can handle fuzzy input vectors as well as real input vectors. In both cases, outputs from the fuzzy neural network are fuzzy vectors. The input-output relation of each unit of the fuzzy neural network is defined by the extension principle of Zadeh. Next we define a cost function for the level sets (i.e., a-cuts) of fuzzy outputs and fuzzy targets. Then we derive a learning algorithm from the cost function for adjusting three parameters of each triangular fuzzy weight. Finally, we illustrate our approach by computer simulations on numerical examples.
In this article, we deal with the one-step time-delay problem in power converter and motor drive system applications implemented by the digital signal processor. For these wide applications, a novel one-step time-dela...
详细信息
In this article, we deal with the one-step time-delay problem in power converter and motor drive system applications implemented by the digital signal processor. For these wide applications, a novel one-step time-delay compensation algorithm is proposed, comprising a standard Luenberger-type observer including a learning algorithm and disturbance observer (DOB). The first novelty is to suggest the learning algorithm automatically determining the observer feedback gains according to the estimation error magnitude. The second is to introduce DOBs with estimation error integrators for removing the steady-state estimation errors, removing the additional adaptive and nonlinear damping compensation terms. The closed-loop observer behaviors are rigorously analyzed to present useful properties. The experimental data obtained using a 3-kW prototype dc-dc boost converter validates the effectiveness of the proposed algorithm.
We study the uniform graph partitioning problem using the learning algorithm proposed by one of us. We discuss the characteristics of the learning algorithm and compare the performance of the algorithm empirically wit...
详细信息
We study the uniform graph partitioning problem using the learning algorithm proposed by one of us. We discuss the characteristics of the learning algorithm and compare the performance of the algorithm empirically with the Kernighan-Lin algorithm on a range of instances. Even with a simple implementation, the learning algorithm is capable of producing very good results.
Training neural networks is a NP complete problem. A learning algorithm is proposed in this paper. Theoretic analysis and simulation results show that this kind of algorithm, which is used to train large BP neural net...
详细信息
ISBN:
(纸本)780003285X
Training neural networks is a NP complete problem. A learning algorithm is proposed in this paper. Theoretic analysis and simulation results show that this kind of algorithm, which is used to train large BP neural network, can speed up the learning rate and gain accurate results.
Artificial neural networks have shown great success in solving real-world problems in recent years. Nowadays, most of the widely used neural network algorithms are running on silicon-based computers, where the resourc...
详细信息
Artificial neural networks have shown great success in solving real-world problems in recent years. Nowadays, most of the widely used neural network algorithms are running on silicon-based computers, where the resource requirement and energy consumption become a challenge when the network size grows. In contrast, brain contains trillions of neurons and synapses which naturally processes information with far less energy as compared to silicon-based computers. In vitro biological neural networks created from dissociated neurons may be used for computing and performing machine learning tasks. In this dissertation, learning algorithms are explored for a network of neurons with detailed biological properties with feedforward and recurrent structures. For feedforward networks, a two-layer hybrid bio-silicon platform is constructed and a five-step design method is proposed for the fast development of living neural network algorithms. Neural variations and dynamics are verified by fitting model parameters with biological experimental results. Random connections are generated under different connection probabilities to vary network sparsity. A multi-layer perceptron algorithm is tested with biological constraints to investigate the impact of neural variations and random connections. The results show that a reasonable inference accuracy can be achieved despite the presence of neural variations and random network connections. A new adaptive pre-processing technique is proposed to ensure good learning accuracy with different living neural network sparsity. On the training side, a supervised STDP-based learning algorithm is proposed for networks with biological constraints. For recurrent spiking neural networks (RSNN), temporal dynamics are studied with detailed biological features. An automatic fitting tool is used to match the precise spike timing of the in vitro neurons and the modeled neurons to get the fitted neuron parameters. Model fidelity and learning performance of dif
暂无评论