Document classification is one of the central issues in information retrieval research. The aim is to uncover similarities between text documents. In other words, classification techniques are used to gain insight in the structure of the various data items contained in the text archive. In this paper we show the results from using a hierarchy of self-organizing maps to perform the text classification task. Each of the individual self-organizing maps is trained independently and gets specialized to a subset of the input data. As a consequence, the choice of this particular artificial neural network model enables the true establishment of a document taxonomy. The benefit of this approach is a straightforward representation of document similarities combined with dramatically reduced training time. In particular, the hierarchical representation of document collections is appealing because it is the underlying organizational principle in use by librarians providing the necessary familiarity...

This paper presents results generated with a new evolutionary robotics (ER) simulation environment and its complementary real mobile robot colony research test-bed. Neural controllers producing mobile robot maze searching and exploration behaviors using binary tactile sensors as inputs were evolved in a simulated environment and subsequently transferred to and tested on real robots in a physical environment. There has been a considerable amount of proof-of-concept and demonstration research done in the field of ER control in recent years, most of which has focused on elementary behaviors such as object avoidance and homing. Artificial neural networks (ANN) are the most commonly used evolvable controller paradigm found in current ER literature. Much of the research reported to date has been restricted to the implementation of very simple behaviors using small ANN controllers. In order to move beyond the proof-of-concept stage our ER research was designed to train larger more complicated ANN controllers, and to implement those controllers on real robots quickly and efficiently. To achieve this a physical robot test-bed that includes a colony of eight real robots with advanced computing and communication abilities was designed and built. The real robot platform has been coupled to a simulation environment that facilitates the direct wireless transfer of evolved neural controllers from simulation to real robots (and vice versa). We believe that it is the simultaneous development of ER computing systems in both the simulated and the physical worlds that will produce advances in mobile robot colony research. Our simulation and training environment development focuses on the definition and training of our new class of ANNs, networks that include multiple hidden layers, and tim...

Document classification may be regarded as one of the central issues in information retrieval research during the last decades. The challenge of classification is to uncover the similarities between groups of data in order to improve the retrieval effectiveness of the overall system. From an exploratory data analysis point of view the same process of classification may be used to gain insight in the structure of the various data items and may thus be referred to as data mining in text archives. In this paper we show the results from applying a neural network model, the hierarchical feature map, to such a data mining task. The neural network is carefully designed to impose a hierarchical structure on the underlying document collection which leads to straight-forward representation of data similarities. Apart from the benefit for text data mining, we are able to demonstrate that the hierarchical feature map leads to a tremendous speed-up of the training process as compared to more tradit...

: In the course of data modeling, many models could be created. Much work has been done on formulating guidelines for model selection. However, by and large, these guidelines are conservative or too specific. Instead of using general guidelines, models could be selected for a particular task based on statistical tests. When selecting one model, others are discarded. Instead of losing potential sources of information, models could be combined to yield better performance. We review the basics of model selection and combination and discuss their differences. Two examples of opportunistic and principled combinations are presented. The first demonstrates that mediocre quality models could be combined to yield significantly better performance. The latter is the main contribution of the paper; it describes and illustrates a novel heuristic approach called the SG(k-NN) ensemble for the generation of good quality and diverse models that can even improve excellent quality models. Key...

In engineering systems, early detection of the occurrence of faults is critical in avoiding product defects. This problematic is here discussed in the framework of an industrial process, namely, an injection moulding plastic machine. The relationships between the process state and the product quality are achieved through Principal Component Analysis. After having identified the main variables, two neural network architectures were investigated, TDNN and Elman networks, with respect to one-step ahead prediction. The results show that TDNN exhibited lower training times with respect to a desired performance criteria. However, for time series in which temporal dependency is large, the recurrent networks with time delayed inputs could lead to better results. 1 Introduction Neural networks have achieved, in recent years, a high degree of importance. The availability of process control computers as well as data historians made it easy to develop neural network solutions for process modeling...

Abstract—In this paper, an optimized approximation algorithm (OAA) is proposed to address the overfitting problem in function approximation using neural networks (NNs). The optimized approximation algorithm avoids overfitting by means of a novel and effective stopping criterion based on the estimation of the signal-to-noise-ratio figure (SNRF). Using SNRF, which checks the goodness-of-fit in the approximation, overfitting can be automatically detected from the training error only without use of a separate validation set. The algorithm has been applied to problems of optimizing the number of hidden neurons in a multilayer perceptron (MLP) and optimizing the number of learning epochs in MLP’s backpropagation training using both synthetic and benchmark data sets. The OAA algorithm can also be utilized in the optimization of other parameters of NNs. In addition, it can be applied to the problem of function approximation using any kind of basis functions, or to the problem of learning model selection when overfitting needs to be considered. Index Terms—Function approximation, neural network (NN) learning, overfitting.

Abstract – This paper presents an artificial intelligence based software tool, entitled the NeuroWizard, which is intended for the engineering design optimization context. The NeuroWizard software tool leverages artificial neural network algorithms as function approximators for empirical modeling of nonlinear dynamic systems or processes, for which closed-form mathematical models either do not exist or are largely inadequate. The user, assumed to be new to the field of artificial intelligence but otherwise well-versed for engineering design optimization, interacts with the software through a wizard that utilizes the graphical user interface within Windows ™ operating system environment. The wizard is embedded with empirical know-how and heuristics from the domain of artificial neural networks to facilitate a relatively effortless deployment of the software tool for the intended user profile. The architecture for the software tool and its projected role within the design optimization framework are detailed.

Large genome sequencing projects generate huge number of protein sequences in their primary structures that is difficult for conventional biological techniques to determine their corresponding 3D structures and then their functions. Protein secondary structure prediction is a prerequisite step in determining the 3D structure of a protein. In this research a method for prediction of protein secondary structure has been proposed and implemented together with other known accurate methods in this domain. The method has been discussed and presented in a comparative analysis progression to allow easy comparison and clear conclusions. A benchmark data set is exploited in training and testing the methods under the same hardware, platforms, and environments. The newly developed method utilizes the knowledge of the GORV information theory and the power of the neural network to classify a novel protein sequence in one of its three secondary structures classes. NN-GORV-I is developed and implemented to predict proteins secondary structure using the biological information conserved in neighboring residues and related

ecial demands of modelling in social science, especially the description of nonlinear quantitative and qualitative relations, stochastic influences, birth and death processes, and micro and multilevel models. The aim is that describing models in MIMOSE should not burden the modeller with a lot of programming and implementation details. MIMOSE was created by Michael M ohring of Computer Science Applications in the Social Sciences, Department of Computer Science, University of Koblenz-Landau, Rheinau 1, D-56075 Koblenz, Germany. Release 2.0 requires Sun Sparc (SunOS, Solaris, X11R5/6), LINUX, or NeXT (Intel/Motorola). A Java interface is under development and the next release will be usable with Java-enabled browsers. SWARM http://www.santafe.edu/projects/swarm/ Swarm is a software package for multi-agent simulation of complex systems being developed at the Santa Fe Institute. It is intended to be a useful tool for researchers in a variety of disciplines, esp

Abstract—This paper presents the research carried out in order to obtain the most efficient neural network design, that allows to approach the steel armours response under high speed projectile impact. One of the main problems related to neural networks is the high quantity of data needed for their training and testing, as well as the complexity of their numeric simulation. In the domain of ballistic impact the large number of parameters involved in the problem hampers the simulation of a large number of impact cases. Trying to solve this issue, the developed research established as aims on the one hand to minimise the number of data used for training, and on the other hand to analyse the influence of the different training variables or input parameters on the learning ability of the created network. The results obtained highlight the small number of data needed to obtain acceptable results, as well as the clear influence of some variables on the generalization ability of the network for this domain.