Neural Network Logic Research Articles

Memetic algorithms, domain knowledge, and financial investing

How might domain knowledge constrain a genetic algorithm and systematically impact the algorithm’s traversal of the search space? In particular, in this paper the hypothesis is advanced that a semantic tree of financial knowledge can be used to influence the results of a genetic algorithm for financial investing problems. An algorithm is described, called the “Memetic Algorithm for Domain Knowledge”, and is instantiated in a software system. In mutation experiments, this system chooses financial ratios to use as inputs to a neural logic network which classifies stocks as likely to increase or decrease in value. The mutation is guided by a semantic tree of financial ratios. In crossover experiments, this system solves a portfolio optimization problem in which components of an individual represent weights on stocks; knowledge in the form of a semantic tree of industries determines the order in which components are sorted in individuals. Both synthetic data and real-world data are used. The experimental results show that knowledge can be used to reach higher fitness individuals more quickly. More interestingly, the results show how conceptual distance in the human knowledge can correspond to distance between evolutionary individuals and their fitness. In other words, knowledge might be dynamically used to at times increase the step size in a search algorithm or at times to decrease the step size. These results shed light on the role of knowledge in evolutionary computation and are part of the larger body of work to delineate how domain knowledge might usefully constrain the genetic algorithm.

A New Control and Design of PEM Fuel Cell System Powered Diffused Air Aeration System

The goal of aquaculture ponds is to maximize production and profits while holding labor and management efforts to the minimum. Poor water quality in most ponds causes risk of fish kills, disease outbreaks which lead to minimization of pond production. Dissolved Oxygen (DO) is considered to be among the most important water quality parameters in fish culture. Fish ponds in aquaculture farms are usually located in remote areas where grid lines are at far distance. Aeration of ponds is required to prevent mortality and to intensify production, especially when feeding is practical, and in warm regions. To increase pond production it is necessary to control dissolved oxygen. Aeration offers the most immediate and practical solution to water quality problems encountered at higher stocking and feeding rates. Many units of aeration system are electrical units so using a continuous, high reliability, affordable, and environmentally friendly power sources is necessary. Fuel cells have become one of the major areas of research in the academia and the industry. Aeration of water by using PEM fuel cell power is not only a new application of the renewable energy, but also, it provides an affordable method to promote biodiversity in stagnant ponds and lakes. This paper presents a new design and control of PEM fuel cell powered a diffused air aeration system for a shrimp farm in Mersa Matruh in Egypt. Also Artificial intelligence (AI) control techniques are used to control the fuel cell output power by controlling its input gases flow rate. Moreover the mathematical modeling and simulation of PEM fuel cell is introduced. A comparative study is applied between the performance of fuzzy logic controller (FLC) and neural network controller (NNC). The results show the effectiveness of NNC over FLC.

Granular neural networks

Fuzzy neural networks (FNNs) and rough neural networks (RNNs) both have been hot research topics in the artificial intelligence in recent years. The former imitates the human brain in dealing with problems, the other takes advantage of rough set theory to process questions uncertainly. The aim of FNNs and RNNs is to process the massive volume of uncertain information, which is widespread applied in our life. This article summarizes the recent research development of FNNs and RNNs (together called granular neural networks). First the fuzzy neuron and rough neuron is introduced; next FNNs are analysed in two categories: normal FNNs and fuzzy logic neural networks; then the RNNs are analysed in the following four aspects: neural networks based on using rough sets in preprocessing information, neural networks based on rough logic, neural networks based on rough neuron and neural networks based on rough-granular; then we give a flow chart of the RNNs processing questions and an application of classical neural networks based on rough sets; next this is compared with FNNs and RNNs and the way to integrate is described; finally some advice is given on development of FNNs and RNNs in future.

Adaptive output feedback tracking controller for a class of uncertain strict feedback nonlinear systems in the absence of state measurements

In this article, design of an adaptive control scheme for a class of uncertain single-input single-output systems in strict feedback form via a backstepping technique has been proposed. It is assumed that system output and its derivatives are available. By virtue of the observability concept, it is shown that for this class of systems there exists a one-to-one map, which maps output and its derivatives to system states. By means of this mapping and using linearly parametrised approximators, such as fuzzy logic systems or neural networks, the uncertain nonlinear dynamics and unavailable states are estimated. The proposed adaptive controller guarantees that the closed-loop system is uniformly ultimately bounded and the influence of minimum approximation error on the L 2-norm of the output tracking error is attenuated arbitrarily. The effectiveness of the proposed scheme has been demonstrated through simulation results.

Recurrent and Feedforward Neural Network Control of Permanent Magnet Synchronous Motor

In this paper, the fuzzy logic, neuro-fuzzy logic, and artificial neural network speed controllers in feedforward and recurrent architectures are designed and modeled to control the permanent magnet synchronous motor (PMSM) drive. The fundamentals of artificial intelligence-based control algorithms are illustrated. The detailed performance of the PMSM drive is studied by applying various mechanical disturbances such as starting performance, speed reversal, flux weakening, and load perturbations.

Neuro-Fuzzy Based Modeling of Electrostatic Fields for Harmattan Season in Zaria

This paper presents a Neuro-Fuzzy based modeling of electrostatic fields for harmattan season in Zaria, Nigeria based on online based data capturing mechanism, which involved the use of a data acquisition system interfaced with a digital electrostatic field strength meter (model 257D) and a computer system. The acquired electric field data were captured by the computer using the Microsoft Office Excel Program for twenty-four months (February 2007 – February 2009). The focus of the analysis is determining the effect of environmental factors such as temperature, pressure and relative humidity on the static electric field during the harmattan season. The plots of the electrostatic field against the variation of the environmental factors were used as the qualitative analytical tools and yielded a non-linear relationship. The data was analyzed using Neuro-Fuzzy technique, which is a hybrid intelligent system combining the benefits of computational techniques of Fuzzy Logic and Artificial Neural Networks. The result of the analyses yielded good neural statistical values of Root Mean Square (RMS) of 0.32, Average Absolute Error of 0.18, and Pearson R value of 0.96 for the harmattan scenario, which are reflections of a good model. The result was further buttressed by the 3D plot of the Neuro-Fuzzy based modeling of the experimental parameters. With the insignificant values of the RMS and Average Absolute value, the empirical model gave a fairly good prediction which could be relied upon to predict the electrostatic fields during harmattan in Zaria, Nigeria.

Fuzzy and Fuzzy Grey-Box Modelling for Entry Temperature Prediction in a Hot Strip Mill

In hot strip mills, initial controller set points have to be calculated before the steel bar enters the mill. Calculations rely on the good knowledge of rolling variables. Measurements are only available once the bar has entered the mill; therefore, they have to be estimated. Estimation of process variables, particularly temperature, is of crucial importance for the bar front section to fulfil quality requirements and must be performed in the shortest possible time to keep heat. Variable estimation is highly affected by measurement uncertainties, variations in the incoming bar conditions, and final product changes. In order to overcome these problems, artificial intelligence techniques, such as fuzzy logic and artificial neural networks, have been proposed. In this article, fuzzy logic-based systems, including fuzzy-based Grey-Box models, are applied to estimate scale breaker entry temperature, given its importance, and its performance is compared against that of the physical model used in plant. Six fuzzy systems and six fuzzy-based Grey-Box models of the type Mamdani, Sugeno, and adaptive neuro-fuzzy inference systems are designed for two different sets of rules and tested with experimental data.

A Method of the Image Emotion Notation Based Fuzzy Reasoning

In this paper, we firstly get the Regions of Interest (ROI) using the Eye tracker and divide every image into two regions including ROI and Non- Regions of Interest (Non-ROI). Secondly, we extract the features of the two regions including ROI and Non-ROI, and get the whole features including texture feature and color feature. Finally, a fuzzy inference network model for image emotion notation using neural logic network is presented. The model describes how to classify the images into different emotions, and neural logic network is used to classification. The learning method proposed is multi-featured, and it allows taking into account the possible predictive power of a simultaneously considered feature conjunction. On the other hand, the feature space partition allows a fuzzy representation of the features and data imprecision integration.

Doubly periodic traveling waves in cellular neural networks with polynomial reactions

In a study (Szekely, 1965) [1] of the locomotion of salamanders, it is observed that a ‘doubly periodic traveling wave solution’ of a logical neural network can be used to explain a dynamic pattern of movements. We show here that nonlinear and nonlogical artificial neural network can also be built by means of reaction diffusion principles and existence or nonexistence of doubly periodic traveling waves can be guaranteed by adjusting parameters built into these networks. Our derivations for existence are based on implicit function theorems and the invariance properties of our model; while nonexistence is based on boundedness properties of the polynomial reaction term. We also give illustrative examples as well as comments on the differences between present results with those obtained for linear models studied earlier in Cheng and Lin (2009) [2,3].

Pattern Recognition Techniques Applied to the Detection and Classification of Welding Defects by Magnetic Testing

The importance of nondestructive tests cannot be denied nowadays, and their use in the petroleum industry is extremely important to ensure safe operation of the equipment. The magnetic flux leakage (MFL) technique in instrumented pig is frequently used for pipeline inspection to detect weld defects and internal or external corrosion; however, as every nondestructive technique, it may present report errors, caused by several human factors (like lack of theoretical knowledge or experience in the technique applied, tiredness due to long hours of work, and so on) and by the inspection technique itself. Because pattern recognition techniques are already well known and applied in other technological research areas, such as fuzzy logic and artificial neural networks, the possibility to develop an automatic system to detect and classify defects through some nondestructive techniques aroused, mainly for those techniques that actually operate with signal or image interpretation. In the present work, pattern nonlinear classifiers were applied using artificial neural networks, to check the possibility to detect and classify defects in pipelines inspected through magnetic pigs (MFL technique). Several tests were performed on specimens with defects artificially inserted (internal and external corrosion and lack of penetration).

Training a Neural Logic Network to predict financial returns: a case study

Companies in the S&P 500 produce quarterly financial statements that are closely studied by investors who forecast the stock market performance of those companies. This paper describes a Neural Logic Network (NLN) for predicting stock market returns based on financial ratios from financial statements. A learning algorithm modifies the NLN in trying to make its output for each stock better correspond with the future market returns of that stock. The input to the NLN is financial ratios taken from quarterly financial statements of companies in the S&P 500. The hypothesis is that the accuracy of financial prediction can be improved by incorporating knowledge into the learning process. This knowledge can be incorporated either as evolutionary knowledge or financial knowledge. Experiments are designed and implemented to test this hypothesis. The results show that evolutionary knowledge improved performance but financial knowledge did not. The specifics of the financial knowledge situation demonstrate some of the challenges of effectively importing financial knowledge into an evolutionary programme.

Fuzzy Logic and Artificial Neural Network approaches for dissolved oxygen prediction

A study on application of data-driven models namely the rule-based model based on mamdani Fuzzy Logic and Artificial Neural Network model in predicting dissolved oxygen in an effluent-impacted urban river is presented and compared. Combined rule bases were formed from the generated fuzzy rules for input – output mapping. Predictability of both the models was good with better performance for the ANN model. With ANN model, it is possible to obtain an output by simulating the model with the parameters of the best-fit model. A relationship connecting the selected parameters not considered previously to assess dissolved oxygen could also be established.

Magnetostrictive micropositioning device with fuzzy-neural-based controller

In precision engineering, adequate performance of ultra precision machine tools depends upon the characteristics of the positioning systems. The purpose of this work is analysis and synthesis to improve the dynamic performance of micropositioning system using magnetostrictive actuators. Various control strategies are considered and evaluated to linearise the actuator's response characteristics of microprocessor-based digital motion controllers. Fuzzy logic (FL) and artificial neural network (ANN) based motion control algorithms have been chosen, due to their advantage of not requiring a complete knowledge of a mathematical model of the controlled mechanism. They are relatively easier to implement and provide a more robust and flexible control system than traditional approaches. Experimental results show that micropositioners driven by magnetostrictive actuators have better dynamics behaviours. This allows the use of such actuators as a valid alternative for positioning in the submicrometer range.

Neuro-fuzzy sliding mode controller: design and stability analysis

In this paper, design and stability analysis of neuro-fuzzy sliding mode controller is discussed. The controller has two parts: fuzzy logic system and neural network. They are used concurrently but each part is responsible for one phase of sliding mode controller. The fuzzy logic system is utilised to control reaching phase dynamics and the feed-forward neural network is employed to keep the system states on the sliding surface. The neural network is trained online using modified back-propagation algorithm. Initially, fuzzy logic system is dominant and as the system moves from reaching phase to sliding phase, neural network becomes more active and hence, a hybrid computing paradigm is achieved. The stability of the system is analysed using Lyapunov's direct method. The proposed controller is implemented to regulate a second-order nonlinear uncertain system and simulation results confirm that the proposed system reduces chattering and improves transient response.

Periodic travelling waves in an artificial neural network

In a study (Szekely, Acta Physiol. Hung. 27 (1965), pp. 285–289) of the locomotion of salamanders, it is observed that a ‘doubly periodic travelling wave solution’ of a logical neural network can be used to explain a dynamic pattern of movements. We show here that a relatively simple (nonlogical) artificial neural network can also be built and necessary and sufficient conditions for the existence of doubly periodic travelling wave solutions can be found. It is hoped that our investigation will set some foundation in the future design of other artificial neural networks that also allow periodic travelling wave solutions.

Rough Neural Network Based on Bottom-Up Fuzzy Rough Data Analysis

Based on bottom-up fuzzy rough data analysis, a new rough neural network decision-making model is proposed. Through supervised Gaustafason---Kessel (G---K) clustering algorithm, proper fuzzy clusters are found to partition the input data space. At the same time cluster number is searched by monotone increasing process. If the cluster number matches with that exactly exist in data sets then excellent fuzzy rough data modeling (FRDM) model can be built. And by integrating it with neural network technique, corresponding rough neural network is constructed. Our method overcomes the defects of conventional top-down based rough logic neural network (RLNN) method, and it also achieves adaptive learning ability and comprehensive soft decision-making ability compared with FRDM model. The experiment results indicate that our method has stronger generalization ability and more compact network structure than conventional RLNN.

New dimensions in non‐classical neural computing, part II: quantum, nano, and optical

PurposeThe purpose of this paper is to introduce new non‐classical implementations of neural networks (NNs). The developed implementations are performed in the quantum, nano, and optical domains to perform the required neural computing. The various implementations of the new NNs utilizing the introduced architectures are presented, and their extensions for the utilization in the non‐classical neural‐systolic networks are also introduced.Design/methodology/approachThe introduced neural circuits utilize recent findings in the quantum, nano, and optical fields to implement the functionality of the basic NN. This includes the techniques of many‐valued quantum computing (MVQC), carbon nanotubes (CNT), and linear optics. The extensions of implementations to non‐classical neural‐systolic networks using the introduced neural‐systolic architectures are also presented.FindingsNovel NN implementations are introduced in this paper. NN implementation using the general scheme of MVQC is presented. The proposed method uses the many‐valued quantum orthonormal computational basis states to implement such computations. Physical implementation of quantum computing (QC) is performed by controlling the potential to yield specific wavefunction as a result of solving the Schrödinger equation that governs the dynamics in the quantum domain. The CNT‐based implementation of logic NNs is also introduced. New implementations of logic NNs are also introduced that utilize new linear optical circuits which use coherent light beams to perform the functionality of the basic logic multiplexer by utilizing the properties of frequency, polarization, and incident angle. The implementations of non‐classical neural‐systolic networks using the introduced quantum, nano, and optical neural architectures are also presented.Originality/valueThe introduced NN implementations form new important directions in the NN realizations using the newly emerging technologies. Since the new quantum and optical implementations have the advantages of very high‐speed and low‐power consumption, and the nano implementation exists in very compact space where CNT‐based field effect transistor switches reliably using much less power than a silicon‐based device, the introduced implementations for non‐classical neural computation are new and interesting for the design in future technologies that require the optimal design specifications of super‐high speed, minimum power consumption, and minimum size, such as in low‐power control of autonomous robots, adiabatic low‐power very‐large‐scale integration circuit design for signal processing applications, QC, and nanotechnology.

A neuro-fuzzy warning system for combating cybersickness in the elderly caused by the virtual environment on a TFT-LCD

Only a few studies in the literature have focused on the effects of age on virtual environment (VE) sickness susceptibility and even less research was carried out focusing on the elderly. In general, the elderly usually browse VEs on a thin film transistor liquid crystal display (TFT-LCD) at home or somewhere, not a head-mounted display (HMD). While the TFT-LCD is used to present VEs, this set-up does not physically enclose the user. Therefore, this study investigated the factors that contribute to cybersickness among the elderly when immersed into a VE on TFT-LCD, including exposure durations, navigation rotating speeds and angle of inclination. Participants were elderly, with an average age of 69.5 years. The results of the first experiment showed that the rate of simulator sickness questionnaire (SSQ) scores increases significantly with navigational rotating speed and duration of exposure. However, the experimental data also showed that the rate of SSQ scores does not increase with the increase in angle of inclination. In applying these findings, the neuro-fuzzy technology was used to develop a neuro-fuzzy cybersickness-warning system integrating fuzzy logic reasoning and neural network learning. The contributing factors were navigational rotating speed and duration of exposure. The results of the second experiment showed that the proposed system can efficiently determine the level of cybersickness based on the associated subjective sickness estimates and combat cybersickness due to long exposure to a VE.

Rough neural network modeling based on fuzzy rough model and its application to texture classification

On the basis of fuzzy rough model (FRM), a method to construct rough neural network FRM_RNN_M is proposed. By means of adaptive Gaustafason–Kessel (G–K) clustering algorithm, fuzzy partition can be accomplished in input feature space. Then based on the search of cluster numbers and attribute subsets, optimal FRM model will be found, and by integrating it with neural network technique, FRM_RNN_M is constructed. The experiment results of classifying Brodatz texture image indicate that FRM_RNN_M is superior to traditional Bayesian and learning vector quantization (LVQ) methods, moreover, FRM_RNN_M has more powerful comprehensive soft decision-making ability than single FRM model. Also, experimental results show that it is favorable to obtain better FRM model if the search of reductive attribute subsets is included. Compared with conventional rough logic neural network (RLNN), FRM_RNN_M has superiorities in the size of structure, convergence speed and generalization ability.

Digital Controller Design Based on Logic Neural Networks

In this paper a new approach on digital controller design based on logic neural networks is presented. In the first part of the paper the logic neuron and its logic function are introduced. Based on this, there is written the general equation for a logic neural network, which allows the easy design of any digital controller directly from the truth table. This solution is validated through MatLab Simulink modeling and simulation and then is implemented in a VLSI chip using PALASM HDL language and also in a ladder diagram PLC device. This method has the advantages of simplicity and flexibility and represents an easy to design modern solution.