Reinforcement Learning Using the Stochastic Fuzzy Min–Max Neural Network

The fuzzy min–max neural network constitutes a neural architecture that is based on hyperbox fuzzy sets and can be incrementally trained by appropriately adjusting the number of hyperboxes and their corresponding volumes. An extension to this network has been proposed recently, that is based on the notion of random hyperboxes and is suitable for reinforcement learning problems with discrete action space. In this work, we elaborate further on the random hyperbox idea and propose the stochastic fuzzy min–max neural network, where each hyperbox is associated with a stochastic learning automaton. Experimental results using the pole balancing problem indicate that the employment of this model as an action selection network in reinforcement learning schemes leads to superior learning performance compared with the traditional approach where the multilayer perceptron is employed.     

AR–ARCH Type Artificial Neural Network for Forecasting

Neural Processing Letters - Real-world time series such as econometric time series are rarely linear and they have characteristics of volatility. Although autoregressive conditional...     

An integrated Data Envelopment Analysis–Artificial Neural Network–Rough Set Algorithm for assessment of personnel efficiency

Personnel specifications have greatest impact on total efficiency. They can help us to design work environment and enhance total efficiency. Determination of critical personnel attributes is a useful procedure to overcome complication associated with multiple inputs and outputs. The proposed algorithm assesses the impact of personnel efficiency attributes on total efficiency through Data Envelopment Analysis (DEA), Artificial Neural Network (ANN) and Rough Set Theory (RST). DEA has two roles in the proposed integrated algorithm of this study. It provides data ANN and finally it selects the best reduct through ANN result. Reduct is described as a minimum subset of attributes, completely discriminating all objects in a data set. The reduct selection is achieved by RST. ANN has two roles in the integrated algorithm. ANN results are basis for selecting the best reduct and it is also used for forecasting total efficiency. The proposed integrated approach is applied to an actual banking system and its superiorities and advantages are discussed.     

NNF and NNPrF—Fuzzy Petri Nets Based on Neural Network for Knowledge Representation, Reasoning and Learning

This paper proposes NNF-a fuzzy Petri Net system based on neural network for proposition logic repesentation,and gives the formal definition of NNF.For the NNF model,forward reasoning algorithm,backward reasoning algorithm and knowledge learning algorithm are discussed based on weight training algorithm of neural network-Back Propagation algorithm.Thus NNF is endowed with the ability of learning a rule.The paper concludes with a discussion on extending NNF to predicate logic,forming NNPrF,and proposing the formal definition and a reasoning algorithm of NNPrF.     

Neural Network State Learning Based Adaptive Terminal ILC for Tracking Iteration-varying Target Points

Terminal iterative learning control(TILC) is developed to reduce the error between system output and a fixed desired point at the terminal end of operation interval over iterations under strictly identical initial conditions. In this work, the initial states are not required to be identical further but can be varying from iteration to iteration. In addition, the desired terminal point is not fixed any more but is allowed to change run-to-run. Consequently, a new adaptive TILC is proposed with a neural network initial state learning mechanism to achieve the learning objective over iterations. The neural network is used to approximate the effect of iteration-varying initial states on the terminal output and the neural network weights are identified iteratively along the iteration axis.A dead-zone scheme is developed such that both learning and adaptation are performed only if the terminal tracking error is outside a designated error bound. It is shown that the proposed approach is able to track run-varying terminal desired points fast with a specified tracking accuracy beyond the initial state variance.     

Attribute Reduction Based on Inclusion Degree for Incomplete and Fuzzy Decision Information Systems

The information systems with incomplete attribute values and fuzzy decisions commonly exist in many applications whose knowledge reduction is one of the most important practical significance. Model of incomplete and fuzzy decision information system is firstly constructed. On the basis of the notion of inclusion degree between fuzzy sets, the attribute reduction for incomplete and fuzzy decision information system, which ensures invariable inclusion degree between every tolerance class and fuzzy decision set, is raised. To reduce the complexity of finding attribute reduction, discernibility sets, discernibifity matrixes and the minimal disjunctive normal form of discernibility sets for incomplete and fuzzy decision information system are introduced. Finally, the algorithm and an example are given, and the solution of the example is proved that the approach to attribute reduction based on inclusion degree is valid.     

A Max–Min ant colony algorithm for fractal dimension of complex networks

The fractal is an important feature of many real complex networks. According to the definition of the Hausdorff dimension, the minimum number of boxes that can be used to cover complex networks is an important factor for revealing the fractal feature of self-similar complex networks. The calculation of the minimum number of boxes is an NP (Non-deterministic Polynomial)-hard problem. In this paper, a heuristic algorithm, named the Max–Min ant colony algorithm, is introduced to approximate the minimum number of boxes. The pheromone-updating rules and heuristic rules are redefined to improve the performance of the algorithm. The experimental results show that, for Escherichia coli networks, the number of boxes was significantly decreased compared to the box-covering greedy algorithm, especially when the box size is small.     

The Neural Network Approach to Input–Output Analysis for Economic Systems

Economic practitioners in China are giving up the classical Leontief’s Input–Output analysis methods. This paper offers an alternative method of input–output analysis. The proposed method is based on the layered neural network model. It shows that neural networks method can be useful for input–output analysis for a dynamic economic system.       

A Max–Min Ant System for the split delivery weighted vehicle routing problem

In real-word cargo transportation practice, such as the deliveries of perishable food and hazardous materials, neglecting the cargo weight in a typical vehicle routing problem (VRP) may prevent the routes from being the most cost effective. Thus, this paper proposes the split-delivery weighted vehicle routing problem (SDWVRP), which consists of constructing the optimal routes, with respect to the constraints on vehicle capacity and cargo weight, to serve a given set of customers with the minimum cost. A Max–Min Ant System (MMAS-SD) algorithm to solve the SDWVRP is developed and a set of theorems and corollaries are proposed to provide an easy approach for route splitting in a typical Weighted VRP (WVRP). The benefit of Split-Delivery for WVRP, as compared to that of SDVRP, primarily lies in its impact on the geographic position and loading weight feature. Large sets of benchmark instances, which are classified into cluster, random and mix types of the three different distribution types, are calculated to demonstrate the effectiveness of the SDWVRP modeling. In addition, the comparison between SDWVRP and WVRP is also carried out via analysis of vehicle numbers, total cost-savings, and the impact of weight variance and mean weight on the ratio of cost-savings and related vehicle numbers of SDWVRP over WVRP to demonstrate the superiority of SDWVRP in determining optimal routes and resulting in substantial cost savings.     

Generating Demand Functions for Data Plans from Mobile Network Operators Based on Users’ Profiles

The evaluation of pricing approaches for mobile data services proposed in the literature can rarely be done in practice. Evaluation by simulation is the most common practice. In these proposals demand and utility functions that describe the reaction of users to offered service prices, use traditional and arbitrary functions (linear, exponential, logit, etc.). In this paper, we present a new approach to construct a simulation model whose output can be used as an alternative method to create demand functions avoiding to use arbitrary and predefined demand functions. However, it is out of the scope of this paper to utilize them to propose pricing approaches, since the main objective of this article is to show the difference between the arbitrary demand functions used and our approach that come from users’ data. The starting point in this paper is to consider data offered from Eurostat, although other data sources could be used for the same purposes with the aim to offer more realistic values that could characterize more appropriately, what users are demanding. In this sense, some demographic and psychographic characteristics of the users are included and others such as the utilization of application usage profiles, as parameters that are included in the user`s profiles. These characteristics and usage profiles make up the user profile that will influence users’ behavior in the model. Using the same procedure, Mobile Network Operators could feed their customers’ data into the model and use it to validate their pricing approaches more accurately before their real implementation or simulate future or hypothetical scenarios. It also makes possible to segment users and make insights for decision-making. Results presented in this paper refer to a simple study case, since the purpose of the paper is to show how the proposal model works and to reveal its differences with arbitrary demand functions used. Of course, results depend on the set of parameters assigned to characterize each user’s profile.     

A combined Independent Component Analysis–Neural Network model for forecasting exchange rate variation

The currency market is one of the most efficient markets, making it very difficult to predict future prices. Several studies have sought to develop more accurate models to predict the future exchange rate by analyzing econometric models, developing artificial intelligence models and combining both through the creation of hybrid models. This paper proposes a hybrid model for forecasting the variations of five exchange rates related to the US Dollar: Euro, British Pound, Japanese Yen, Swiss Franc and Canadian Dollar. The proposed model uses Independent Component Analysis (ICA) to deconstruct the series into independent components as well as neural networks (NN) to predict each component. This method differentiates this study from previous works where ICA has been used to extract the noise of time series or used to obtain explanatory variables that are then used in forecasting. The proposed model is then compared to random walk, autoregressive and conditional variance models, neural networks, recurrent neural networks and long–short term memory neural networks. The hypothesis of this study supposes that first deconstructing the exchange rate series and then predicting it separately would produce better forecasts than traditional models. By using the mean squared error and mean absolute percentage error as a measures of performance and Model Confidence Sets to statistically test the superiority of the proposed model, our results showed that this model outperformed the other models examined and significantly improved the accuracy of forecasts. These findings support this model’s use in future research and in decision-making related to investments.     

Flexible Knowledge–Vision–Integration Platform for Personal Protective Equipment Detection and Classification Using Hierarchical Convolutional Neural Networks and Active Leaning

ABSTRACT

This work is part of an effort to develop of a knowledge–vision integration platform for hazard control in industrial workplaces, adaptable to a wide range of industrial environments. The paper focuses on hazards resulted from the nonuse of personal protective equipment. The objective is to test the capability of the platform to adapt to different industrial environments by simulating the process of randomly selecting experiences from a new scenario, querying the user, and using their feedback to retrain the system through a hierarchical recognition structure using convolutional neural network (CNN). Thereafter, in contrast to the random sampling, the concept of active learning based on pruning of redundant points is tested. Results obtained from both random sampling and active learning are compared with a rigid systems that is not capable to aggregate new experiences as it runs. From the results obtained, it can be concluded that the classification accuracy improves greatly by adding new experiences, which makes it possible to customize the service according to each scenario and application as it functions. In addition, the active learning approach was able to reduce the user query and slightly improve the overall classification performance, when compared with random sampling.     

Improved Accuracy Function for Interval-Valued Intuitionistic Fuzzy Sets and Its Application to Multi–Attributes Group Decision Making

Accuracy functions proposed by various researchers fail to compare some interval-valued intuitionistic fuzzy sets (IVIFSs) correctly. In the present research paper, we propose an improved accuracy function to compare all comparable IVIFSs correctly. The use of proposed accuracy function is also proposed in a method for multi attribute group decision making (MAGDM) method with partially known attributes’ weight. Finally, the proposed MAGDM method is implemented on a real case study of evaluation teachers’ performance. Sensitivity analysis of this method is also done to show the effectiveness of the proposed accuracy function in MAGDM.     

Existence and Exponential Stability of Periodic Solution for BAM Fuzzy Cohen–Grossberg Neural Networks with Mixed Delays

In this paper, BAM fuzzy Cohen–Grossberg neural networks with mixed delays are considered. Using M-matrix theory and differential inequality techniques, some sufficient conditions for the existence and exponential stability of periodic solution to the neural networks are established. The results of this paper are completely new and complementary to the previously known results. Finally, an example is given to illustrate the effectiveness of our results obtained.     

An optimum feature extraction method based on Wavelet–Radon Transform and Dynamic Neural Network for pavement distress classification

Quantification of pavement crack data is one of the most important criteria in determining optimum pavement maintenance strategies. Recently, multi-resolution analysis such as wavelet decompositions provides very good multi-resolution analytical tools for different scales of pavement analysis and distresses classification. This paper present an automatic diagnosis system for detecting and classification pavement crack distress based on Wavelet–Radon Transform (WR) and Dynamic Neural Network (DNN) threshold selection. The algorithm of the proposed system consists of a combination of feature extraction using WR and classification using the neural network technique. The proposed WR + DNN system performance is compared with static neural network (SNN). In test stage; proposed method was applied to the pavement images database to evaluate the system performance. The correct classification rate (CCR) of proposed system is over 99%. This research demonstrated that the WR + DNN method can be used efficiently for fast automatic pavement distress detection and classification. The details of the image processing technique and the characteristic of system are also described in this paper.     

Periodic Solution for Fuzzy Cohen–Grossberg BAM Neural Networks with Both Time-Varying and Distributed Delays and Variable Coefficients

In this paper, we investigate the existence and global exponential stability of periodic solution for a general class of fuzzy Cohen–Grossberg bidirectional associative memory (BAM) neural networks with both time-varying and (finite or infinite) distributed delays and variable coefficients. Some novel sufficient conditions for ascertaining the existence, uniqueness, global attractivity and exponential stability of the periodic solution to the considered system are obtained by applying matrix theory, inequality analysis technique and contraction mapping principle. The results remove the usual assumption that the activation functions are bounded and/or continuously differentiable. It is believed that these results are significant and useful for the design and applications of fuzzy Cohen–Grossberg BAM neural networks. Moreover, an example is employed to illustrate the effectiveness and feasibility of the results obtained here.     

Design of Delay-Dependent Exponential Estimator for T–S Fuzzy Neural Networks with Mixed Time-Varying Interval Delays Using Hybrid Taguchi-Genetic Algorithm

This paper considers the design of state estimator for Takagi?CSugeno (T?CS) fuzzy neural networks with mixed time-varying interval delays. The mixed time-delays that consist of both the discrete time-varying and distributed time-delays with a given range are presented. The activation functions are assumed to be globally Lipschitz continuous. By using the Lyapunov-Krasovskii method, a linear matrix inequality (LMI) approach is developed to construct sufficient conditions for the existence of admissible state estimator such that the error-state system is exponentially globally stable. To avoid complex mathematical derivations and conservative results, a new hybrid Taguchi-genetic algorithm method is integrated with a LMI method to seek the estimator gains that satisfy the Lyapunov-Krasovskii functional stability inequalities. The proposed new approach is straightforward and well adapted to the computer implementation. Therefore, the computational complexity can be reduced remarkably and facilitate the design task of the estimator for T?CS fuzzy neural networks with time-varying interval delays. Two illustrative examples are exploited in order to illustrate the effectiveness of the proposed state estimator.     

Takagi–Sugeno Fuzzy Hopfield Neural Networks for $${\mathcal{H}_{\infty}}$$ Nonlinear System Identification

In this paper, we propose a new H_￥{\mathcal H_\infty} weight learning algorithm (HWLA) for nonlinear system identification via Takagi–Sugeno (T–S) fuzzy Hopfield neural networks with time-delay. Based on Lyapunov stability theory, for the first time, the HWLA for nonlinear system identification is presented to reduce the effect of disturbance to an H_￥{\mathcal{H}_{\infty }} norm constraint. The HWLA can be obtained by solving a convex optimization problem which is represented in terms of linear matrix inequality (LMI). An illustrative example is given to demonstrate the effectiveness of the proposed identification scheme.