Aleksander Malinowski

Sensitivity analysis for minimization of input data dimension for feedforward neural network

Multilayer feedforward networks are often used for modeling complex relationships between the data sets. Deleting unimportant data components in the training sets could lead to smaller networks and reduced-size data vectors. This can be achieved by analyzing the total disturbance of network outputs due to perturbed inputs. The search for redundant data components is performed for networks with continuous outputs and is based on the concept in sensitivity of linearized neural networks. The formalized criteria and algorithm for pruning data vectors are formulated and illustrated with examples.<<ETX>>

Efficient algorithm for training neural networks with one hidden layer

Efficient second order algorithm for training feedforward neural networks is presented. The algorithm has a similar convergence rate as the Lavenberg-Marquardt (LM) method and it is less computationally intensive and requires less memory. This is especially important for large neural networks where the LM algorithm becomes impractical. Algorithm was verified with several examples.

Comparative Study of Derivative Free Optimization Algorithms

Derivative free optimization algorithms are often used when it is difficult to find function derivatives, or if finding such derivatives are time consuming. The Nelder Meads simplex method is one of the most popular derivative free optimization algorithms in the fields of engineering, statistics, and sciences. This algorithm is favored and widely used because of its fast convergence and simplicity. The simplex method converges really well with small scale problems of some variables. However, it does not have much success with large-scale problems of multiple variables. This factor has reduced its popularity in optimization sciences significantly. Two solutions of quasi-gradients are introduced to improve it in terms of the convergence rate and the convergence speed. The improved algorithm with higher success rate and faster convergence which still maintains the simplicity is the key feature of this paper. This algorithm will be compared on several benchmark functions with other popular optimization algorithms such as the genetic algorithm, the differential evolution algorithm, the particle swarm algorithm, and the original simplex method. Then, the comparing results will be reported and discussed.

Inverse control of nonlinear systems using neural network observer and inverse mapping approach

This paper introduces a new approach to inverse control. Unlike using commonly known method of plant inverse dynamics learning, the control sequence is calculated using inverse mapping approach. Both methods are compared with nonlinear plant examples and selected desired waveforms.

Internet as a new graphical user interface for the SPICE circuit simulator

The Spice Internet Package (SIP) was developed using an Internet browser as a platform-independent graphical user interface. The SIP application has many options that include simulation of SPICE files, graphical postprocessing data, and online editing of SPICE files. It can run remotely through a network on any operating system.

Guest editorial: Distributed data processing in industrial applications

The papers in this special section focus on the deployment of distributed data processing in industrial applications. The main objective is both bringing together ideas of the worldwide research community about one common platform and presenting the latest advances and developments in design, modeling, programming, management, and innovative implementations of distributed information systems, including the latest works related to industrial communication technologies to be used in the future.

Automated Data Mining from Web Servers Using Perl Script

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 required data. The resultant HTML file consisting of the required data is displayed for the perusal of users. Future enhancements to our ethernet robot include optimization to improve performance and customization for use as a sophisticated client-specific search agent.

Internet based neural network online simulation tool

Since the legendary work of McCulloch and Pitts from early 1940s and introduction of concept of artificial neuron, numerous attempts aiming to automate the process of training neural networks have been made. Neural networks, even though successfully applied in many different areas still bear significant problems with respect to adequate choice of network parameters, architecture, etc. As the concept of neural networks evolved through more than six decades, the technology also underwent tremendous changes. Educational tools aiming to help everyday users to learn, gain and apply knowledge in neural network training have evolved accordingly. Unlike many commercial or freeware tools available on market today, authors have decided to go with the implementation that would provide two significant advantages to an existing tools. In this paper authors are proposing an educational too that is characterized by both transparent accessibility with respect to hardware and software platform on one, and ease of use on the other hand. The tool is web based, therefore user is relieved of installation and setup hurdles. Besides, by employing a remote server that hosts the application, the user saves local resources for other jobs.

Environments for rapid implementation of control algorithms and hardware-in-the-loop simulation

Integrated control design environments reduce the time of development, simulation and real-time testing of embedded control systems. This tutorial paper describes five contemporary software and hardware tools which may be used to create integrated control design environments to facilitate rapid implementation of control algorithms and hardware-in-the-loop simulation. Examples presented include, decoupling stabilizing control of an unstable multivariable plant, digital PID control of an analog plant and PWM control of DC motor speed.

SIP-Spice Intranet Package

Computer network communications have become a popular and efficient means of computing. Most companies and research institutions use networks to some extent on a regular basis. This paper demonstrates the capability of computer-aided design (CAD) through computer networks. There are several benefits that make networked CAD tools desirable. A networked application can be used remotely through any network connection. Any operating system can be used to access a networked application, making the application operating system independent. Also the application can be made to run on a pay-per-use basis if licensing is desired, and much less installation time and configuration time is required because the application is located on one central machine. The Spice Intranet Package (SIP) is demonstrated as an example of such a networked CAD application. This particular application allows users to run Spice simulations and view graphical analysis of electronic circuits through a network connection using a Web browser. The SIP application has many options that include simulation and analysis of Spice files, graphical analysis of data, online editing of Spice files, passwords with separate file areas for each user, and a user friendly graphical user interface.