Learning methods of constructive neural network aims to overcome the disadvantages of BP algorithm, which has many advantages such as fast convergent rate, less computation, good fault-tolerant and strong generalizati...
详细信息
ISBN:
(纸本)9781615677214
Learning methods of constructive neural network aims to overcome the disadvantages of BP algorithm, which has many advantages such as fast convergent rate, less computation, good fault-tolerant and strong generalization ability, etc. Its structure is constructed step by step in the processing of the data rather than being prescribed in advance. This paper mainly introduces the developing motivations, research status quo and development directions, laying stress on two kinds of constructive learning methods: FP algorithm and covering algorithm. After discussing the construction and basic properties of FP network, we construct an FP network as a general clustering unit, and analyze its main properties. On the base of summarizing covering algorithm we present a general neighborhood covering algorithm and its corresponding network. According to the characteristics of the algorithm we analyze its existing problems and propose the further research orientations.
In this paper, we use the relations of quotient space theory and martingale therory to research the iterated function system that is fractal geometry images, and propose these conclusions: Given an irreducible iterate...
详细信息
ISBN:
(纸本)7900719229
In this paper, we use the relations of quotient space theory and martingale therory to research the iterated function system that is fractal geometry images, and propose these conclusions: Given an irreducible iterated function system {X,wipij;i,j = 1,2,}, then exists a corresponding chain of quotient space {Wk = (Xk, μk,Fk);&=1,2,:} and a martingale {(μk, Fk );k = 1,2,}on the chain, therefore there are: l)Assume Pk is a invariant subsets of Wk, P is a invariant subsets of W, then exists lim k&rarr∞ Pk =P and the convergence is according to Hausdorff distance. 2)Assume μk is a invariant measure of Fk, μ is a invariant measure of F, then exists limμk k&rarr∞=μ 3) Pk is a support set of μk, P is a support set of μ Namely we present the quotient approximation theorem about fractal geometry images, and build relations among chain of quotient space, martingale, fractal geometry images and Markovian process.
THz wave generated by difference-frequency in nonlinear crystal in normal temperature is theoretically studied. The conclifton of phase matching is deduced and an experimental setup is proposed. With this method, THz ...
详细信息
THz wave generated by difference-frequency in nonlinear crystal in normal temperature is theoretically studied. The conclifton of phase matching is deduced and an experimental setup is proposed. With this method, THz wave can be generated without rigorous condition and can work continuously for a long period of time in room temperature.
The DS-CDMA signal model and the noisy linear independent component analysis (ICA) model are analyzed in this paper. Comparing these models shows that they have the same form. The adaptive minimum mean-square error (M...
The DS-CDMA signal model and the noisy linear independent component analysis (ICA) model are analyzed in this paper. Comparing these models shows that they have the same form. The adaptive minimum mean-square error (MMSE) multiuse detection based on ICA is proposed. It uses the output of adaptive MMSE multi-user detection to initialize the ICA iterations, not only the known spread information of interesting user is used to overcome the uncertainness of ICA, but also the character of statistical independence is used. The simulation results show that the performance is improved obviously.
Engineering optimization in the intelligence swarm remains to be a challenge. Recently, a novel optimization method based on number-theory and particle swarm, good lattice swarm optimization algorithm(GLSO), is introd...
详细信息
Engineering optimization in the intelligence swarm remains to be a challenge. Recently, a novel optimization method based on number-theory and particle swarm, good lattice swarm optimization algorithm(GLSO), is introduced, which intends to produce faster and better global search ability and more accurate convergence because it has a solid theoretical basis. In this paper, four models of constructing good point set are introduced and the GLSO based on new models is rewritten. Some applications of the new model on constrained engineering via employing a penalty function approach suggest that the presented algorithm is potentially a powerful search technique for solving complex engineering design optimization problems.
The mechanism of the classical particle swarm optimization and the comparison criterion of different natural computing methods is investigated by introducing the discrepancy and good lattice points in number theory an...
详细信息
The mechanism of the classical particle swarm optimization and the comparison criterion of different natural computing methods is investigated by introducing the discrepancy and good lattice points in number theory and proposes a novel optimization method, called good lattice points-based particle swarm optimization algorithm, which intends to produce faster and more accurate convergence because it has a solid theoretical basis and better global search ability, meanwhile the global convergence of the presented algorithm with asymptotic probability one is proved by the property of the optimal lattice. Finally experiment results are very promising to illustrate the outstanding feature of the presented algorithm.
In this article, we propose a (t,n) threshold verifiable multi-secret sharing scheme, in which to reconstruct t secrets needs to solve t simultaneous equations. The analysis results show that our scheme is as easy as ...
详细信息
In this article, we propose a (t,n) threshold verifiable multi-secret sharing scheme, in which to reconstruct t secrets needs to solve t simultaneous equations. The analysis results show that our scheme is as easy as Yang's scheme [8] in the secret reconstruction and requires less public values than Chien's [7] and Yang's schemes. Furthermore, the shares in our scheme can be verified their validity with t public values based on ECDLP, and there are two verified forms: one is computationally secure as Feldman 's scheme [12] and other is unconditionally secure as Pedersen's scheme [13]. In addition, for the main computation: a i,1 P 1 + a i,2 P 2 + hellip + a i,t P t in our scheme, we present a new method based on the signed factorial expansion and implement it, the results show that it is more efficient than the current public methods. Thus our scheme is a secure and efficient (t,n) threshold verified multi-secret sharing scheme.
A novel Pareto-based multi-objective fully-informed particle swarm algorithm (FIPS) is proposed to solve flexible job-shop problems in this paper. Firstly, the population is ranked based on Pareto optimal concept. And...
详细信息
A novel Pareto-based multi-objective fully-informed particle swarm algorithm (FIPS) is proposed to solve flexible job-shop problems in this paper. Firstly, the population is ranked based on Pareto optimal concept. And the neighborhood topology used in FIPS is based on the Pareto rank. Secondly, the crowding distance of individuals is computed in the same Pareto level for the secondary rank. Thirdly, addressing the problem of trapping into the local optimal, the mutation operators based on the coding mechanism are introduced into our algorithm. Finally, the performance of the proposed algorithm is demonstrated by applying it to several benchmark instances and comparing the experimental results.
A novel image content authentication algorithm based on Laplace spectra was proposed. Outstanding feature points are extracted from the original image and a cipher point is inserted. A relational graph is then built, ...
详细信息
ISBN:
(纸本)7900719229
A novel image content authentication algorithm based on Laplace spectra was proposed. Outstanding feature points are extracted from the original image and a cipher point is inserted. A relational graph is then built, and the Laplace spectra of the graph are calculated to serve as image features. The Laplace spectra are quantized then embedded into the original image as a watermark. In the authentication step, the Laplace spectra of the authenticating image are calculated and compared with that of the watermark embedded in the authenticating image. If both of the spectra are identical, the image passes the authentication test. Otherwise, the tamper is found. The experimental results show that the proposed authentication algorithm can effectively detect the event and the location when the original image content is tampered viciously.
暂无评论