Bogdan Wilamowski

Auburn University | AU · Department of Electrical & Computer Engineering

Non-intrusive load monitoring (NILM) is to estimate individual appliance’s power consumption from aggregated smart meter data, which is useful for optimized energy management and provisioning of customized services. While deep learning (DL) has achieved state-of-the-art NILM performance, it is still constrained by the dependency on large amounts of...

In this study, a modular design methodology inherited from cognitive neuroscience and neurophysiology is proposed to develop artificial neural networks, aiming to realize the powerful capability of brain—divide and conquer—when tackling complex problems. First, a density-based brain-like partition method is developed to construct the modular archit...

The Fully Connected Cascade Networks (FCCN) were originally proposed along with the Cascade Correlation (CasCor) learning algorithm that having three main advantages over the Multilayer Perceptron (MLP): the structure of the network could be determined dynamically; they were more powerful for complex feature representation; the training was efficie...

This paper proposes a novel unsupervised multi-dimensional scaling (MDS) method to visualize high-dimensional data and their relations in a low-dimensional (e.g., 2D) space. Different from traditional MDS approaches where the main purpose is to embed high-dimensional data into a low-dimensional space, this study aims to both embed data into a low-d...

Radial basis function (RBF) networks, because of their universal approximation ability, have been widely applied to industrial process modeling. In this study, an Improved ErrCor (IErrCor) algorithm is proposed, in which compact structure and satisfactory generalization ability can be obtained with only one learning try. First, a second-order-based...

The discrete cosine transform (DCT) is commonly known in signal processing. In this paper DCT is used in computational intelligence to show its usefulness. Proposed DCT method is used to reduce the size of system which results in faster processing with limited and controlled precision lost. Proposed method is compared to other ones like Fuzzy Syste...

Number of training patterns has a huge impact on artificial neural networks training process, not only because of time-consuming aspects but also on network capacities. During training process the error for the most patterns reaches low error very fast and is hold to the end of training so can be safely removed without prejudice to further training...

Research in deep neural networks are becoming popular in artificial intelligence. Main reason for training difficulties is the problem of vanishing gradients while number of layers increases. While such networks are very powerful they are difficult in training. The paper discusses capabilities of different neural network architectures and presents...

Stress effects in semiconductor devices have gained significant attention in semiconductor industry in recent years, and numerical modeling is often used as a powerful tool for stress analysis in semiconductor devices. Here, we present a nontraditional 1D model for fast stress analysis in bipolar junction transistors. Because bipolar transistors ar...

Currently very popular trend in artificial intelligence is the use of deep neural networks. The power of such networks are very large, but the main difficulty is learning these networks. The article presents a analysis of deep neural network nonlinearity with polynomial approximation of neuron activation functions. It is shown that nonlinearity gro...

According to the simplicity and universal approximation capability, single layer feedforward networks (SLFN) are widely used in classification and regression problems. The paper presents a new OLS-PSO constructive algorithm based on Orthogonal Least Square (OLS) method and Particle Swarm Optimization (PSO) algorithm. Instead of evaluating the ortho...

Traditional bipolar differential amplifiers have only a ±5-mV operational range with nonlinear distortion below 0.1 dB. In this paper, a linearization technique based on neural-network training algorithm is proposed to expand this 0.1-dB linear region to a much wider ±200-mV range. Compared with the traditional and recent state-of-the-art technique...

Traditional data processing algorithms are usually not capable to process big data. As matter of fact, usually big data is being defined as such which cannot be processed with traditional techniques. At the same time a progress of technology makes that humans are now overwhelm by big data. One way of processing big data is to use deep neural networ...

Residual levels of stress remaining after device fabrication have been characterized in the base and emitter regions of shallow-trench-isolated complementary npn and pnp transistors on (100) silicon utilizing uniaxial stress measurements. A residual biaxial stress of approximately 160 MPa has been found in the active regions of the npn transistors,...

Although discovery of the Error Back Propagation (EBP) learning algorithm was a real breakthrough, this is not only a very slow algorithm, but it also is not capable of training networks with super compact architecture. The most noticeable progress was done with an adaptation of the LM algorithm to neural network training. The LM algorithm is capab...

This paper presents the adaptive analog hardware implementation of a MLP (multilayer perceptron architecture) ANN (artificial neural networks) for online nonlinear system operation. Neurons are implemented by bipolar differential pairs with tangent hyperbolic activation function. A bipolar current multiplier and a linearized differential amplifier...

Approximation of unknown functions in multiple dimensions is an important topic in many areas of industrial engineering, such as nonlinear control. Currently, approaches such as neural networks or fuzzy systems are used to create highly nonlinear surfaces from data. Here we show the capabilities of a very simple classical numerical method such as c...

The macroscopic stress dependence of bipolar junction transistors (BJTs) can be modeled by three transport model parameters as a function of stress: saturation current IS, forward current gain βF, and Early voltage VA. Recent research has shown that Early voltage VAis independent of stress, so it is not discussed in detail in this paper. Unfortunat...

Radial basis function(RBF) networks have been proved to be a universal approximator when enough hidden nodes are given and proper parameters are selected. Conventional algorithms for RBF networks training, including two-stage methods and gradient-based algorithms, cost much computation and have difficulty to determine the network size. In this pape...

By introducing an extra dimension to the inputs, sigmoid function can simulate the behavior of traditional RBF units. This paper introduces a sigmoid based RBF neuron and compares it with traditional RBF neuron. Neural networks composed of these neurons are trained with ErrCor algorithm on two classic experiments. Comparison results are presented t...

Magnetic Resonance Imaging (MRI) results in overall quality that usually calls for human intervention in order to correctly identify details present in the image. More recently, interest has arisen in automated processes that can adequately segment medical image structures into substructures with finer detail than other efforts to-date. Relatively...

Transconductance filters use LC ladder filters as prototypes. Unfortunately real LC filters and transconductance filters have losses and the resultant filter characteristics are often very different than the desired one. The purpose of this study is to redesign filters in such way that, despite of nonideal elements, the filter characteristics is ve...

The 29 articles in this special section focus on the topic of intelligent systems.

One of the major difficulties facing researchers using neural networks is the selection of the proper size and topology of the networks. The problem is even more complex because often when the neural network is trained to very small errors, it may not respond properly for patterns not used in the training process. A partial solution proposed to thi...

This paper proposes an improved second order (ISO) algorithm for training radial basis function (RBF) networks. Besides the traditional parameters, including centers, widths and output weights, the input weights on the connections between input layer and hidden layer are also adjusted during the training process. More accurate results can be obtain...

Radial basis function (RBF) networks have advantages of easy design, good generalization, strong tolerance to input noise, and online learning ability. The properties of RBF networks make it very suitable to design flexible control systems. This paper presents a review on different approaches of designing and training RBF networks. The recently dev...

The paper presents the state-of-the-art study on the recently published literatures subjected to industrial control in various industrial applications. Controllers are classified into several types according to different control technologies, including PID algorithm, Kalman filter, least squares regression, network-based automation, fuzzy inference...

The quality of journals is primarily related to the number of citation of published papers. Various measures of evaluating the quality of journals are being discussed and compared here. Unfortunately the citation analysis is almost impossible using manual examination of references. This must be done by developing special computer tools for extracti...

The number of computer based functions embedded in vehicles has increased significantly in the past two decades. An in-vehicle embedded electronic architecture is a complex distributed system; the development of which is a cooperative work involving different manufacturers and suppliers. There are several key demands in the development process, suc...

The paper presents the state-of-the-art study on the recently published literatures subjected to power aware design in various industrial applications. The basic conception of power-aware design is placed at first and then the recent progress is investigated. For each paper, a brief summary is given to introduce the related power awareness design t...

The paper presents the properties of two types of neural networks: traditional neural networks and radial basis function (RBF) networks, both of which are considered as universal approximators. In this paper, the advantages and disadvantages of the two types of neural network architectures are analyzed and compared based on four different examples....

Soft computing can be a very attractive alternative to a purely digital system, but there are many traps waiting for researchers trying to apply this new exciting technology. For nonlinear processing both neural networks and fuzzy systems can be used. Terrifically neural networks should provide much better solutions: smoother surfaces, larger numbe...

Humans are not perfect and they make mistakes. For example, humans without computer aided tools would not be able to design VLSI chips larger than 100 transistors. Computers are assisting humans in many aspects of their life. For many years already computers were used for number crunching, office related jobs, etc. More recently computers are used...

The paper presents the design of three types of neural networks with different features, including traditional backpropagation networks, radial basis function networks and counterpropagation networks. Traditional backpropagation networks require very complex training process before being applied for classification or approximation. Radial basis fun...

Neural network research over the past 3 decades has resulted in improved designs and more efficient training methods. In today's high-tech world, many complex non-linear systems described by dozens of differential equations are being replaced with powerful neural networks, making neural networks increasingly more important. However, all of the curr...

In the presentation major difficulties of designing neural networks are shown. It turn out that popular MLP (Multi Layer Perceptron) networks in most cases produces far from satisfactory results. Also, popular EBP (Error Back Propagation) algorithm is very slow and often is not capable to train best neural network architectures. Very powerful and f...

This paper describes a method of linearizing the nonlinear characteristics of many sensors and devices using an embedded neural network. The neuron-by-neuron process was developed in assembly language to allow the fastest and shortest code on the embedded system. The embedded neural network also requires an accurate approximation for hyperbolic tan...

This paper introduces a generalized method for simulating dynamic systems in SPICE. The proposed method is useful for students in that the necessary software is free and only an elementary knowledge of SPICE is required. The method uses a netlist description of the system that comprises little more than a set of state equations. In comparison with...

The method introduced in this paper allows for training arbitrarily connected neural networks, therefore, more powerful neural network architectures with connections across layers can be efficiently trained. The proposed method also simplifies neural network training, by using the forward-only computation instead of the traditionally used forward a...

Humans are very capable of solving many scientific and engineering problems, but during the solution process they have a tendency to make mistakes. For example, humans without computer aided tools, would not be able to design VLSI chips larger than 100 transistors. This imperfection of humans make them very unreliable elements in resilient control...

The presentation is focused on comparison of neural networks and fuzzy systems. Advantages and disadvantages of both technologies are discussed. Fuzzy systems are relatively easy to design but number of inputs in the system are significantly limited. It is very difficult to design neural networks so rather they have to be trained instead. Neural ne...

With the dramatic increase of network bandwidth and decrease of network latency and because of development of new network programming technologies, the dynamic websites provide dynamic interaction to the end user and at the same time implement asynchronous client-server communication in the background. Many applications are being deployed through t...

In this paper, a neural architecture which gives identical TSK fuzzy system is proposed based on the area selection concept in neural network design. Instead of using traditional membership functions for selection the range of operation, the monotonic pair-wire or sigmoidal activation function is used. In the comparison to popular neuro-fuzzy syste...

The improved computation presented in this paper is aimed to optimize the neural networks learning process using Levenberg-Marquardt (LM) algorithm. Quasi-Hessian matrix and gradient vector are computed directly, without Jacobian matrix multiplication and storage. The memory limitation problem for LM training is solved. Considering the symmetry of...

This paper describes a method of linearizing the nonlinear characteristics of many sensors using an embedded neural network. The proposed method allows for complex neural networks with very powerful architectures to be embedded on a very inexpensive 8-bit microcontroller. In order to accomplish this unique training software was developed as well as...

This paper introduces a neural network training tool through computer networks. The following algorithms, such as neuron by neuron (NBN) , error back propagation (EBP), Levenberg Marquardt (LM) and its improved versions are implemented in two different computing methods, traditional forward-backward computation and newly developed forward-only comp...

With the increase of Internet bandwidth the World Wide Web is changing the approach for software development. Traditionally, most of the software was developed for one particular platform such as DOS, Windows, Mac, Unix or Linux and it was not portable from one system to another. Usually the user interface is system dependent and it has to be devel...

Neural networks are the topic of this paper. Neural networks are very powerful as nonlinear signal processors, but obtained results are often far from satisfactory. The purpose of this article is to evaluate the reasons for these frustrations and show how to make these neural networks successful. The following are the main challenges of neural netw...

The paper describes basic concepts of neural networks and fuzzy systems. It is shows that most commonly used neural network architecture of MLP – Multi Layer Perceptron is also one of the least efficient ones. Also most commonly used EBP – Error Back Propagation algorithm is not only very slow, but also it is not able to find solutions for optimal...

This paper introduces a neural network training tool, NBN 2.0, which is developed based on neuron by neuron computing method. Error backpropagation (EBP) algorithm, Levenberg Marquardt (LM) algorithm and its improved versions are implemented in two different computing methods, traditional forward-backward computation and newly developed forward-onl...

The comparisons of various learning algorithms were presented and it was shown that most popular neural network topologies (MLP) and most popular training algorithm (EBP) are not giving optimal solution. Instead MLP networks much simpler neural network topologies can be used to produce similar or better results. Instead of popular EBP more advance...

In the paper, issues concerning the synthesis of ladder reactance analog filter have been raised. In all existing algorithms of synthesis of such filters it is assumed that the reactance elements are ideal ones. Since actual reactance elements are lossy, which means the Q-factor has a finite value, the frequency response of real filters differs wid...

Article data extraction from internet is a way to download and extract the required data automatically from web servers. In this paper, we present a method called the Internet Robot to extract the data directly from a web server by using Perl scripting language with the powerful regular expressions. The regular expressions are widely used in this m...

The paper is going to introduce a revised C++ version of neural network trainer (NNT) which is developed based on neuron by neuron computation. Besides traditional error back propagation (EBP) algorithm, two improved version of Levenberg Marquardt (LM) algorithm and a newly developing algorithm are also implemented. The software can handle not only...

An overview of various neural network architectures is presented. Depending on applications some of these architectures are capable to perform very complex operations with limited number of neurons, while other architectures, which use more neurons, are easy to train. There are neural network architectures which have very limited requirements for t...

In this paper, a new tunable CMOS resistor is proposed. The resistance is inversely proportional to bias current, to provide the resistor with a wider tuning range. And transistors, composing the active resistor, work at saturation region to achieve very large resistance within a small area. As an example, a low pass RC filter using tunable CMOS re...

Fabrication of patterned diamond structures in an inexpensive way is desirable for a variety of practical applications. Inkjet printing is a well-developed and inexpensive process by which liquid ink, as well as solid suspensions in a properly formulated solution, can be applied in a precise quantity and at selected locations on a rigid or flexible...

This paper describes a new algorithm with neuron-by-neuron computation methods for the gradient vector and the Jacobian matrix. The algorithm can handle networks with arbitrarily connected neurons. The training speed is comparable with the Levenberg-Marquardt algorithm, which is currently considered by many as the fastest algorithm for neural netwo...

Composite thin films of nanodiamond and silica nanotubes were synthesized by means of microwave plasma assisted chemical vapor deposition (MPCVD) on silica nanotube matrix that was seeded with nanodiamond particles. SEM, Raman spectroscopy, and EDX were used to analyze the composite. Wet chemical etching was applied to selectively remove exposed si...

An optimization algorithm is presented which effectively combines the desirable characteristics of both gradient descent and evolutionary computation into a single robust algorithm. The method uses a population-based gradient approximation which allows it to recognize surface behavior on both large and small scales. By adjusting the population radi...

The paper presents a review of electronic noses with emphasis of the usage of live olfactory receptor neurons as detectors interfaced with electronics. The paper focuses on the pattern recognition issue using artificial neural networks. The proposed architecture seems to be very simple and powerful at the same time. The architecture was verified in...

Electroencephalogram signals (EEG), also known as brainwaves, provide rich information about the brain processes, once that they result from the electrical activity of millions of neurons beneath the skull. By finding correlation between EEG patterns and certain brain processes, such as thought, it is possible to design an intelligent system that i...

This work is concerned with the design automation of analog circuits realizing piecewise linear functions (PWL) that may be used for fuzzy logic circuit design. There are several sources of systematic or random errors in the design of such functions. Various combinations of CMOS current mirror circuits are used to implement PWL functions. In order...

There are significant efforts to develop gyroscopes using MEMS technology; accuracies of gyroscopes varying from rate-grade, through tactical-grade, to inertial grade. The random walk varies from 0.5 °/&surd;h through 0.05 °/&surd;h to 0.001 °/&surd;h. The most common approach is to use vibratory gyros, which use mechanical elements (proof-mass) to...

Not Available

The paper describes a multiobjective optimization method which combines a second order algorithm with an enhanced evolutionary search in order to obtain a set of points which lie on the Pareto-optimal front. The second order portion of the method makes use of a "quasi-Jacobian" modification of the Levenberg-Marquardt algorithm, while the evolutiona...

The paper describes a neural network implementation on a low end and inexpensive microcontroller. It also describes the method of using a simple hardware multiplier to generate multibyte accurate results. An activation function that is equivalent to tangent hyperbolic is also described. An example is shown using an inexpensive eight bit microcontro...

Data mining from the Web is the process of extracting essential data from any web server. In this paper, we present a method called Ethernet Robot to extract information/data from a web server using perl scripting language and to process the data using regular expressions. The procedure involves fetching, filtering, processing and presentation of r...

Fuzzy controllers are easy to design for complex control surfaces but produce rough control surfaces which might lead to unstable operation. On the other hand neural controllers are hard and complex to train but they produce very accurate output control surfaces compared to that of fuzzy controllers. The neuro-fuzzy controller proposed in this pape...

The paper shows that it fully connected neural networks are used then the same problem can be solved with less number of neurons and weights. Interestingly such networks are trained faster. The problem is that most of the neural networks terming algorithms are not suitable for such network. Presented algorithm and software allow training feedforwad...

The paper describes an optimization method which combines advantages of both evolutionary computation and gradient based methods. The proposed method follows the general concept of evolutionary computation, but uses an approximated gradient for generating subsequent populations. The gradient is not explicitly computed, but is instead estimated usin...

Generation of filters using operational transconductance amplifiers and capacitors (OTA-C) has become more and more popular. A simple yet effective method of analytical synthesis is presented that makes use of ladder circuits, which are much more widely understood. Given the ladder circuit of a filter of any order, an equivalent OTA-C filter can be...

Although neural networks have been around for over 20 years, we still have difficulties training them. Training is often difficult and time consuming. The paper describes a software (NNT) developed for neural network training. In addition to the traditional Error Back Propagation (EBP) algorithm, several second order algorithms were implemented. Th...

Nonlinear processes are difficult to control because there can be so many variations of the nonlinear behavior. The issue becomes more complicated if a nonlinear characteristic of the system changes with time and there is a need for an adaptive change of the nonlinear behavior. These adaptive systems are best handled with methods of computational i...

Not Available

In this paper, a model of LAN traffic is presented. In the model, the most important components influencing the network traffic are taken into account. Namely, the transmission protocols and information buffering, operating systems, and queuing algorithms as well as user behavior in network applications are considered. The model is based on an "on-...

Analog to digital converters (ADC) are one of the most critical blocks in the area of electronics. This paper presents two new circuits for Gray-code current-mode analog to digital conversion which employs absolute value operation. The implementations of two different 5-bit converter circuits are discussed in this paper. 0.25 mum CMOS technology wi...

We survey the recent methods for improving position control of electrostatic microelectromechanical (MEMS) actuators, and present recent new nonlinear approaches being studied in our electronic packaging and control systems laboratories. The new methods include a type of variable structure control (switching, but not sliding mode) and two methods f...

The paper presents algorithms which could improve the transmission band for computer networks. The analysis of file size distribution has shown that, for a typical Web site, more than 40% of files are smaller than 1 kB. Additionally, these small files are more frequently used because 80% of all references by a client browser to Web site resources a...

In the paper a model of the traffic in the LAN is presented. In the model the most important components influencing the network traffic are taken into account. Namely, the transmission protocols and information buffering, operational systems and queuing algorithms as well as users' behavior working with the network applications are considered. The...

Digital controllers have historically enjoyed many advantages over those synthesized by analog electronics, but there are still some drawbacks to discrete-time implementations of controllers and signal processing algorithms: costly conversion of analog signals to digital and back, quantization errors, digital noise, time discretization, computation...

MOSFETs used in space are subject to exposure to natural radiation in space. Among the effects of ionizing radiation are shifts in threshold voltage and reduction of carrier mobility. In this paper, total-dose effects in switching dc/dc boost converter are examined using SPlCE simulations. Then a new circuit design for an open loop dc/dc boost conv...

Comparison of various methods of computational intelligence are presented and illustrated with examples. These methods include neural networks, fuzzy systems, and evolutionary computation. The presentation is focused on neural networks, their learning algorithms and special architectures. General learning rule as a function of the incoming signals...