Optimal ensemble construction via meta-evolutionary ensembles

In this paper, we propose a meta-evolutionary approach to improve on the performance of individual classifiers. In the proposed system, individual classifiers evolve, competing to correctly classify test points, and are given extra rewards for getting difficult points right. Ensembles consisting of multiple classifiers also compete for member classifiers, and are rewarded based on their predictive performance. In this way we aim to build small-sized optimal ensembles rather than form large-sized ensembles of individually-optimized classifiers. Experimental results on 15 data sets suggest that our algorithms can generate ensembles that are more effective than single classifiers and traditional ensemble methods.     

Wavefront reduction using graphs,neural networks and genetic algorithm

This paper presents an algorithm for nodal numbering in order to obtain a small wavefront. Element clique graphs are employed as the mathematical models of finite element meshes. A priority function containing five vectors is used, which can be viewed as a generalization of Sloan's function. These vectors represent different connectivity properties of the graph models. Unlike Sloan's algorithm, which uses two fixed coefficients, here, five coefficients are employed, based on an evaluation by artificial neural networks. The networks weights are obtained using a simple genetic algorithm. Examples are included to illustrate the performance of the present hybrid method. Copyright © 2004 John Wiley & Sons, Ltd.     

神经网络在缺陷识别中的应用

为了准确、快速地进行缺陷识别，介绍了一种新型的前馈神经网络模型，即径向基概率神经网络。与以往的算法相比，该方法具有分类识别精度高且速度快的优点。仿真获得了很好的结果。     

A new distributed approximation algorithm for constructing minimum connected dominating set in wireless ad hoc networks

In recent years, constructing a virtual backbone by nodes in a connected dominating set (CDS) has been proposed to improve the performance of ad hoc wireless networks. In general, a dominating set satisfies that every vertex in the graph is either in the set or adjacent to a vertex in the set. A CDS is a dominating set that also induces a connected sub‐graph. However, finding the minimum connected dominating set (MCDS) is a well‐known NP‐hard problem in graph theory. Approximation algorithms for MCDS have been proposed in the literature. Most of these algorithms suffer from a poor approximation ratio, and from high time complexity and message complexity. In this paper, we present a new distributed approximation algorithm that constructs a MCDS for wireless ad hoc networks based on a maximal independent set (MIS). Our algorithm, which is fully localized, has a constant approximation ratio, and O(n) time and O(n) message complexity. In this algorithm, each node only requires the knowledge of its one‐hop neighbours and there is only one shortest path connecting two dominators that are at most three hops away. We not only give theoretical performance analysis for our algorithm, but also conduct extensive simulation to compare our algorithm with other algorithms in the literature. Simulation results and theoretical analysis show that our algorithm has better efficiency and performance than others. Copyright © 2005 John Wiley & Sons, Ltd.     

Interval based MINLP superstructure synthesis of mass exchange networks

A new simultaneous synthesis approach for mass exchange networks (MENs) is presented. The technique is adapted from the interval based mixed integer non-linear program (MINLP) superstructure (IBMS) synthesis of heat exchanger networks (HENs). The superstructure interval boundary compositions are defined by the supply and target compositions of either the rich or lean set of streams. Each rich and lean stream in the superstructure has the potential of exchanging mass with streams of opposite kind in each interval based on mass transfer feasibility. The model harnesses the strengths of the stagewise superstructure and the pinch technology methods for mass exchange network synthesis (MENS). The IBMS approach simultaneously trades-off the capital and operating costs for MENs. The superstructure composition interval defining approach introduced in this paper enforces the mixing of split streams at equal compositions, hence there is no need to include mixing equations in the model. Fixing the interval boundaries helps to eliminate the complexities involved in initialisations, thus the region of search for the optimum solution is reduced. The IBMS model is applied to MENS problems involving continuous contact and staged columns. It is also extended to problems involving multiple mass separating agents (MSAs) and regeneration. The results obtained compare well with those in the literature.     

Stochastic molecular descriptors for polymers. 4. Study of complex mixtures with topological indices of mass spectra spiral and star networks: The blood proteome case

The Quantitative Structure-Property Relationships (QSPRs) based on Graph or Network Theory are important for predicting the properties of polymeric systems. In the three previous papers of this series (Polymer 45 (2004) 3845-3853; Polymer 46 (2005) 2791-2798; and Polymer 46 (2005) 6461-6473) we focused on the uses of molecular graph parameters called topological indices (TIs) to link the structure of polymers with their biological properties. However, there has been little effort to extend these TIs to the study of complex mixtures of artificial polymers or biopolymers such as nucleic acids and proteins. In this sense, Blood Proteome (BP) is one of the most important and complex mixtures containing protein polymers. For instance, outcomes obtained by Mass Spectrometry (MS) analysis of BP are very useful for the early detection of diseases and drug-induced toxicities. Here, we use two Spiral and Star Network representations of the MS outcomes and defined a new type of TIs. The new TIs introduced here are the spectral moments (π_k) of the stochastic matrix associated to the Spiral graph and describe non-linear relationships between the different regions of the MS characteristic of BP. We used the MARCH-INSIDE approach to calculate the π_k(SN) of different BP samples and S2SNet to determine several Star graph TIs. In the second step, we develop the corresponding Quantitative Proteome-Property Relationship (QPPR) models using the Linear Discriminant Analysis (LDA). QPPRs are the analogues of QSPRs in the case of complex biopolymer mixtures. Specifically, the new QPPRs derived here may be used to detect drug-induced cardiac toxicities from BP samples. Different Machine Learning classification algorithms were used to fit the QPPRs based on π_k(SN), showing J48 decision tree classifier to have the best performance. These results suggest that the present approach captures important features of the complex biopolymers mixtures and opens new opportunities to the application of the idea supporting classic QSPRs in polymer sciences.     

Battery-Aware Routing for Streaming Data Transmissions in Wireless Sensor Networks

Recent technological advances have made it possible to support long lifetime and large volume streaming data transmissions in sensor networks. A major challenge is to maximize the lifetime of battery-powered sensors to support such transmissions. Battery, as the power provider of the sensors, therefore emerges as the key factor for achieving high performance in such applications. Recent study in battery technology reveals that the behavior of battery discharging is more complex than we used to think. Battery powered sensors might waste a huge amount of energy if we do not carefully schedule and budget their discharging. In this paper we study the effect of battery behavior on routing for streaming data transmissions in wireless sensor networks. We first give an on-line computable energy model to mathematically model battery discharge behavior. We show that the model can capture and describe battery behavior accurately at low computational complexity and thus is suitable for on-line battery capacity computation. Based on this battery model we then present a battery-aware routing (BAR) protocol to schedule the routing in wireless sensor networks. The routing protocol is sensitive to the battery status of routing nodes and avoids energy loss. We use the battery data from actual sensors to evaluate the performance of our protocol. The results show that the battery-aware protocol proposed in this paper performs well and can save a significant amount of energy compared to existing routing protocols for streaming data transmissions. Network lifetime is also prolonged with maximum data throughput. As far as we know, this is the first work considering battery-awareness with an accurate analytical on-line computable battery model in sensor network routing. We believe the battery model can be used to explore other energy efficient schemes for wireless networks as well.     

Dynamic Multicast Session Provisioning in WDM Optical Networks with Sparse Splitting Capability

In this work, we study dynamic provisioning of multicast sessions in a wavelength-routed sparse splitting capable WDM network with an arbitrary mesh topology where the network consists of nodes with full, partial, or no wavelength conversion capabilities and a node can be a tap-and-continue (TaC) node or a splitting and delivery (SaD) node. The objectives are to minimize the network resources in terms of wavelength-links used by each session and to reduce the multicast session blocking probability. The problem is to route the multicast session from each source to the members of every multicast session, and to assign an appropriate wavelength to each link used by the session. We propose an efficient online algorithm for dynamic multicast session provisioning. To evaluate the proposed algorithm, we apply the integer linear programming (ILP) optimization tool on a per multicast session basis to solve off-line the optimal routing and wavelength assignment given a multicast session and the current network topology as well as its residual network resource information. We formulate the per session multicast routing and wavelength assignment problem as an ILP. With this ILP formulation, the multicast session blocking probability or success probability can then be estimated based on solving a series of ILPs off-line. We have evaluated the effectiveness of the proposed online algorithm via simulation in terms of session blocking probability and network resources used by a session. Simulation results indicate that our proposed computationally efficient online algorithm performs well even when a fraction of the nodes are SaD nodes.     

Moral value transfer from regulatory fit: What feels right is right and what feels wrong is wrong.

People experience regulatory fit (E. T. Higgins, 2000) when the strategic manner of their goal pursuit suits their regulatory orientation, and this regulatory fit feels right. Fit violation feels wrong. Four studies tested the proposal that experiences of fit can transfer to moral evaluations. The authors examined transfer of feeling wrong from fit violation by having participants in a promotion or prevention focus recall transgressions of commission or omission (Studies 1 and 2). Both studies found that when the type of transgression was a fit violation, participants expressed more guilt. Studies 3 and 4 examined transfer of feeling right from regulatory fit. Participants evaluated conflict resolutions (Study 3) and public policies (Study 4) as more right when the means pursued had fit. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

A novel distributed approach to robust fault detection and identification

In this paper, we propose a novel distributed robust fault detection and identification (RFDI) scheme for a class of nonlinear systems. Firstly, a detection and identification estimator—robust fault tracking approximator (RFTA) is designed for online health monitoring. A novel feature of the RFTA is that it can simultaneously detect and accurately identify the shape and magnitude of the fault and disturbance. Moreover, it takes less online training time compared with the traditional neural network based fault diagnosis schemes. For some distributed systems, a network of distributed estimators is constructed where the RFTA is embedded into each estimator. Then we use consensus filter to filter the outputs of each estimator. One of the most important merits of the consensus filter is that its outputs can dramatically improve the accuracy of fault detection and identification. Next, the stability of the distributed RFDI scheme is rigorously investigated. Finally, two numerical examples are given to illustrate the feasibility and effectiveness of the proposed approach.