Search

Automatic image annotation aims to predict labels for images according to their semantic contents and has become a research focus in computer vision, as it helps people to edit, retrieve and understand large image collections. In the last decades, researchers have proposed many approaches to solve this task and achieved remarkable performance on several standard image datasets. In this paper, we propose a novel learning to rank approach to address image auto-annotation problem. Unlike typical learning to rank algorithms for image auto-annotation which directly rank annotations for image, our approach consists of two phases. In the first phase, neural ranking models are trained to rank image’s semantic neighbors. Then nearest-neighbor based models propagate annotations from these semantic neighbors to the image. Thus our approach integrates learning to rank algorithms and nearest-neighbor based models, including TagProp and 2PKNN, and inherits their advantages. Experimental results show that our method achieves better or comparable performance compared with the state-of-the-art methods on four challenging benchmarks including Corel5K, ESP Games, IAPR TC-12 and NUS-WIDE. [ABSTRACT FROM AUTHOR]

This paper examines a linked simulation-optimization procedure based on the combined application of an artificial neural network (ANN) and genetic algorithm (GA) with the aim of developing an efficient model for the multiobjective management of groundwater lenses in small islands. The simulation-optimization methodology is applied to a real aquifer in Kish Island of the Persian Gulf to determine the optimal groundwater-extraction while protecting the freshwater lens from seawater intrusion. The initial simulations are based on the application of SUTRA, a variable-density groundwater numerical model. The numerical model parameters are calibrated through automated parameter estimation. To make the optimization process computationally feasible, the numerical model is subsequently replaced by a trained ANN model as an approximate simulator. Even with a moderate number of input data sets based on the numerical simulations, the ANN metamodel can be efficiently trained. The ANN model is subsequently linked with GA to identify the nondominated or Pareto-optimal solutions. To provide flexibility in the implementation of the management plan, the model is built upon optimizing extraction from a number of zones instead of point-well locations. Two issues are of particular interest to the research reported in this paper are: (1) how the general idea of minimizing seawater intrusion can be effectively represented by objective functions within the framework of the simulation-optimization paradigm, and (2) the implications of applying the methodology to a real-world small-island groundwater lens. Four different models have been compared within the framework of multiobjective optimization, including (1) minimization of maximum salinity at observation wells, (2) minimization of the root mean square (RMS) change in concentrations over the planning period, (3) minimization of the arithmetic mean, and (4) minimization of the trimmed arithmetic mean of concentration in the observation wells. The latter model can provide a more effective framework to incorporate the general objective of minimizing seawater intrusion. This paper shows that integration of the latest innovative tools can provide the ability to solve complex real-world optimization problems in an effective way. [ABSTRACT FROM AUTHOR]

This paper proposes a novel steam valving controller using radial basis function (RBF) networks-based approximate model method. Approximate model method is a kind of direct linearization approach that is derived based on the approximation of the plant's input-output model via Taylor expansion. RBF networks are used to identify the plant to implement the approximate model control law. In order to improve the performance of the approximate model controller, RBF networks weights are adjusted online using BP algorithms with an adaptive learning rate. Several simulations results demonstrate the effectiveness of the proposed controller for team valving control. [ABSTRACT FROM AUTHOR]

In this paper, neural algorithms, including the multi-layered perceptron (MLP) differential approximator, generalized hybrid power series, discrete Hopfield neural network, and the hybrid numerical, are used for constructing models that incorporate a priori knowledge in the form of differential equations for dynamic engineering processes. The properties of these approaches are discussed and compared to each other in terms of efficiency and accuracy. The presented algorithms have a number of advantages over other traditional mesh-based methods such as reduction of the computational cost, speed up of the execution time, and data integration with the a priori knowledge. Furthermore, the presented techniques are applicable when the differential equations governing a system or dynamic engineering process are not fully understood. The proposed algorithms learn to compute the unknown or free parameters of the equation from observations of the process behavior, hence a more precise theoretical description of the process is obtained. Additionally, there will be no need to solve the differential equation each time the free parameters change. The parallel nature of the approaches outlined in this paper make them attractive for parallel implementation in dynamic engineering processes. [ABSTRACT FROM AUTHOR]

Developing a model based digital human meridian system is one of the interesting ways of understanding and improving acupuncture treatment, safety analysis for acupuncture operation, doctor training, or treatment scheme evaluation. In accomplishing this task, how to construct a proper model to describe the behavior of human meridian systems is one of the very important issues. From experiments, it has been found that the hysteresis phenomenon occurs in the relations between stimulation input and the corresponding response of meridian systems. Therefore, the modeling of hysteresis in a human meridian system is an unavoidable task for the construction of model based digital human meridian systems. As hysteresis is a nonsmooth, nonlinear and dynamic system with a multi-valued mapping, the conventional identification method is difficult to be employed to model its behavior directly. In this paper, a neural network based identification method of hysteresis occurring in human meridian systems is presented. In this modeling scheme, an expanded input space is constructed to transform the multi-valued mapping of hysteresis into a one-to-one mapping. For this purpose, a modified hysteretic operator is proposed to handle the extremum-missing problem. Then, based on the constructed expanded input space with the modified hysteretic operator, the so-called Extreme Learning Machine (ELM) neural network is utilized to model hysteresis inherent in human meridian systems. As hysteresis in meridian system is a dynamic system, a dynamic ELM neural network is developed. In the proposed dynamic ELM neural network, the output state of each hidden neuron is fed back to its own input to describe the dynamic behavior of hysteresis. The training of the recurrent ELM neural network is based on the least-squares algorithm with QR decomposition. [ABSTRACT FROM AUTHOR]

Retrieval schemes often use two important components: 1) a radiative transfer model (RTM) inside the retrieval procedure or to construct the learning dataset for the training of the statistical retrieval algorithms and 2) a numerical weather prediction (NWP) model to provide a first guess or, again, to construct a learning dataset. This is particularly true in operational centers. As a consequence, any physical retrieval or similar method is limited by inaccuracies in the RTM and NWP models on which it is based. In this paper, a method for partially compensating for these errors as part of the sensor calibration is presented and evaluated. In general, RTM/NWP errors are minimized as best as possible prior to the training of the retrieval method, and then tolerated. The proposed method reduces these unknown and generally nonlinear residual errors by training a separate preprocessing neural network (NN) to produce calibrated radiances from real satellite data that approximate those radiances produced by the ''flawed'' NWP and RTM models. The final ''compensated/flawed'' retrieval assures better internal consistency of the retrieval procedure and then produces more accurate results. To the authors'' knowledge, this type of NN model has not been used yet for this purpose. The calibration approach is illustrated here on one particular application: the retrieval of atmospheric water vapor from the Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E) and the Humidity Sounder for Brazil (HSB) measurements for nonprecipitating scenes, over land and ocean. Before being inverted, the real observations are ''projected'' into the space of the RTM simulation space from which the retrieval is designed. Validation of results is performed with radiosonde measurements and NWP analysis departures. This study shows that the NN calibration of the AMSR-E/HSB observations improves water vapor inversion, over ocean and land, for both clear and cloudy situations. The NN calibration is efficient and very general, being applicable to a large variety of problems. The nonlinearity of the NN allows for the calibration procedure to be state dependent and adaptable to specific cases (e.g., the same correction will not be applied to medium-range measurement and to extreme conditions). Its multivariate nature allows for a full exploitation of the complex correlation structure among the instrument channels, making the calibration of each single channel more robust. The procedure would make it possible to project the satellite observations in a reference observational space defined by radiosonde measurements, RTM simulations, or other instrument observational space. [ABSTRACT FROM AUTHOR]

This paper presents a neural network algorithm called SOLAR (Supervised One-shot Learning Algorithm for Real-valued inputs) that can complete learning by a single presentation of the real-valued inputs. The similarity matrix is introduced in SOLAR, which can measure the Euclidean distance between training instances represented by a real number. Based on the geometrical structure of the similarity matrix, the structure of the network, the connection weights, the learning parameters and the linear disciminant function are determined. Then, the learning is executed by a single presentation of the training set. The nonlinearly separable case by decomposing the set of supervisor signals into linearly separable sets. The main contribution of this paper is that SOLAR realizes a high-speed learning by a single presentation of the input data composed of real values and a large-scale complex set of data can be learned. This has been difficult in previous models. In addition, since the number of hidden units and the parameters are determined by the algorithm, there is no problem such as trial and error. [ABSTRACT FROM AUTHOR]

This paper proposes an algorithm for creating rules of a fuzzy logic system using a neuro-fuzzy approach. The proposal is based on the results of Juang and Tsao who use a Fuzzy Neural Network (FNN) to generate rules and fuzzy sets from input data. A time series example is solved using our proposal, which is tested using statistical analysis of the residuals of the model. [ABSTRACT FROM AUTHOR]

Abstract: Achieving a low Noise-to-Signal Ratio (NSR) is one of the major concerns when implementing hardware-based neural networks. Continuous Valued Number System (CVNS) features have been exploited to improve the NSR. The efficiency of the network model in terms of area, power consumption, and NSR is measured using the product of the total number of neurons in the network multiplied by the network NSR, which indicates the number of neurons required for a specific NSR. The network proposed in this paper stores the weights in digital registers while the processing is done in the analog domain using CVNS arithmetic. The mathematical analysis and comparison between the proposed network and the previous structures prove that the proposed architecture improves upon in terms of both the NSR and the product of neuron multiplied by NSR. [Copyright &y& Elsevier]

Abstract: This paper uses an artificial neural network (ANN) and Levenberg–Marquardt training algorithm to model the nonlinear relationship between refractive index and energy gap in semiconductors. The predicted simulation values of the ANN are in accordance with the experimental data. An error deviation (Δn) was estimated for different models. The lowest deviation is given by the ANN model. The ANN model performance was also tested for some compounds not included in the training and was found to be in good agreement with the experimental data. High precision of the NN model is demonstrated as well as good generalization performance. [Copyright &y& Elsevier]

Abstract: Neural network systems are highly capable of deriving knowledge from complex data, and they are used to extract patterns and trends which are otherwise hidden in many applications. Preserving the privacy of sensitive data and individuals'' information is a major challenge in many of these applications. One of the most popular algorithms in neural network learning systems is the back-propagation (BP) algorithm, which is designed for single-layer and multi-layer models and can be applied to continuous data and differentiable activation functions. Another recently introduced learning technique is the extreme learning machine (ELM) algorithm. Although it works only on single-layer models, ELM can out-perform the BP algorithm by reducing the communication required between parties in the learning phase. In this paper, we present new privacy-preserving protocols for both the BP and ELM algorithms when data is horizontally and vertically partitioned among several parties. These new protocols, which preserve the privacy of both the input data and the constructed learning model, can be applied to online incoming records and/or batch learning. Furthermore, the final model is securely shared among all parties, who can use it jointly to predict the corresponding output for their target data. [Copyright &y& Elsevier]

Abstract: This paper presents a rule-based approach for the classification of power quality disturbances. The disturbed signal is first characterized using the multi-resolution S-transform, which acts as a feature extraction tool. Then, a simple but robust rule-based classification algorithm is used to identify disturbances. This algorithm uses linear and parabolic rules as pattern classifiers where decision boundaries are established by a heuristic search. The classification algorithm has a modular structure where each module works separately to detect specific disturbances. The most common types of disturbances, including sags, interruptions, swells, harmonics and oscillatory transients, were analyzed. Moreover, complex disturbances consisting of combinations of two simple events (simultaneous or consecutive in the same interval) were also analyzed. In both cases, noise, ranging from 40 to 20dB, was also considered. The tested data set contains power quality signals obtained using mathematical models, power quality events obtained from power network''s simulations using PSCAD/EMTDC and measured signals at electrical installations. Finally, evaluation results verifying the accuracy of the proposed method are presented and compared to those obtained from a classification system based on an artificial neural network. [Copyright &y& Elsevier]

Abstract: Dynamic pricing is a pricing strategy where price for the product changes according to the expected demand for it. Some work on using neural network for dynamic pricing have been previously reported, such as for forecasting the demand and modelling consumer choices. However, little work has been done in using them for optimising pricing policies. In this paper, we describe how neural networks and evolutionary algorithms can be combined together to optimise pricing policies. Particularly, we build a neural network based demand model and use evolutionary algorithms to optimise policy over build model. There are two key benefits of this approach. Use of neural network makes it flexible enough to model a range of different demand scenarios occurring within different products and services, and the use of evolutionary algorithm makes it versatile enough to solve very complex models. We also evaluate the pricing policies found by neural network based model to that found by other widely used demand models. Our results show that proposed model is more consistent, adapts well in a range of different scenarios, and in general, finds more accurate pricing policy than other three compared models. [Copyright &y& Elsevier]

Abstract: In the paper, as an improvement of fuzzy clustering neural network FCNN proposed by Zhang et al., a novel robust fuzzy clustering neural network RFCNN is presented to cope with the sensitive issue of clustering when outliers exist. This new algorithm is based on Vapnik''s ɛ-insensitive loss function and quadratic programming optimization. Our experimental results demonstrate that RFCNN has much better robustness for outliers than FCNN. [Copyright &y& Elsevier]

The inherent black-box nature of neural networks is an important drawback with respect to the problem of explanation of neural network responses. Although several articles have tackled the problem of rule extraction from a single neural network, just a few papers have investigated rule extraction from several combined neural networks. In this article we describe how to translate symbolic rules into the Discretized Interpretable Multi-Layer Perceptron (DIMLP) and how to extract rules from one or several combined neural networks. Our approach consists of characterizing discriminant hyperplane frontiers. Unordered rules are extracted in polynomial time with respect to the size of the problem and the size of the network. Moreover, the degree of matching between extracted rules and neural network responses is 100% on training examples. We applied single DIMLP networks to 17 data sets related to medical diagnosis and medical prognosis problems. Results based on 10-fold cross-validation showed that the DIMLP model was on average as accurate as standard multi-layer perceptrons (MLP). Furthermore, DIMLP networks were significantly more accurate than CN2 on eight problems, whereas only on one problem CN2 was better than DIMLP. Finally, a non-Hodgkin lymphoma diagnosis problem based on classification of electrophoresis gels was defined. It turned out that ensembles of DIMLP networks were significantly more accurate than CN2 (96.1%±1.4 versus 82.7%±4.0). Finally, symbolic rules revealed the presence of five important spots for the discrimination of the class of Lymphocyte Leukemia/Chronic Lymphoid Leukemia (Lc/LLc), and the class of Centrocytic Lymphoma (Cc). [Copyright &y& Elsevier]

This article discusses the identification of nonlinear dynamic systems using multi-layer perceptrons (MLPs). It focuses on both structure uncertainty and parameter uncertainty, which have been widely explored in the literature of nonlinear system identification. The main contribution is that an integrated analytic framework is proposed for automated neural network structure selection, parameter identification and hysteresis network switching with guaranteed neural identification performance. First, an automated network structure selection procedure is proposed within a fixed time interval for a given network construction criterion. Then, the network parameter updating algorithm is proposed with guaranteed bounded identification error. To cope with structure uncertainty, a hysteresis strategy is proposed to enable neural identifier switching with guaranteed network performance along the switching process. Both theoretic analysis and a simulation example show the efficacy of the proposed method. [ABSTRACT FROM AUTHOR]