Wyniki wyszukiwania - BazTech

1

The sigma-if neural network as a method of dynamic selection of decision subspaces for medical reasoning systems

Huk M.

Journal of Medical Informatics & Technologies

|

2004

|

Vol. 7

KB65--73

EN

To-date research in the area of applied medical artificial intelligence systems suggests that it is necessary to focus further on the characteristic requirements of this research field. One of those requirements is related to the need for effective analysis of multidimensional heterogeneous data sets, which poses particular difficulties when considering AI-suggested solutions. Recent works point to the possibility of extending the activation function of a perception to the time domain, thus significantly enhancing the capabilities of neural networks. This change results in the ability to dynamically tune the size of the decision space under consideration, which stems from continuous adaptation of the interneuron connection architecture to the data being classified. Such adaptation reflects the importance of individual decision attributes for the patterns being classified, as defined by the Sigma-if network during its training phase. These characteristics enable effective employment of such networks in solving classification problems, which emerge in medical sciences. The described approach is also a novel, interesting area of neural network research. This article discusses selected aspects of construction as well as training of Sigma-if networks, based on a sample problem of classifying Arabic numeral images.

2

Problems of medical data mining

Szpunar-Huk E.

Journal of Medical Informatics & Technologies

|

2004

|

Vol. 7

KB91--98

EN

The article discusses the main problems connected to the specificity of medical aspects, especially as concerns the quality and means of selection of data and tools used for constructing classification systems. Special attention is devoted to the risks inherent in direct application of classical knowledge extraction algorithms (such as the algorithms for constructing decision trees) to medical data. The article describes some attempts at solving emerging problems and points to the need for analysis of classifiers with regard to more than just their potential redundancy and mutual exclusion. The article also proposes two functions, useful for analysing rule sets with focus on data semantics.

3

Rough Modeling---a Bottom-up Approach to Model Construction

Loken T., Komorowski J.

International Journal of Applied Mathematics and Computer Science

|

2001

|

Vol. 11, no 3

675-690

EN

Traditional data mining methods based on rough set theory focus on extracting models which are good at classifying unseen objects. If one wants to uncover new knowledge from the data, the model must have a high descriptive quality---it must describe the data set in a clear and concise manner, without sacrificing classification performance. Rough modeling, introduced by Kowalczyk (1998), is an approach which aims at providing models with good predictive and descriptive qualities, in addition to being computationally simple enough to handle large data sets. As rough models are flexible in nature and simple to generate, it is possible to generate a large number of models and search through them for the best model. Initial experiments confirm that the drop in performance of rough models compared to models induced using traditional rough set methods is slight at worst, and the gain in descriptive quality is very large.

4

Juicer - a data mining approach to information extraction from the WWW

Masłowska I., Weiss D.

Foundations of Computing and Decision Sciences

|

2000

|

Vol. 25, No. 2

67-87

EN

We present a novel approach to automatic text mining on the World Wide Web. Considering the fact that the enormously dynamic growth of the WWW results in a need for new, more powerful information extraction tools we designed and implemented a system, which adapts techniques originally introduced in the field of data mining. We believe that similar systems, which usually base on machine learning or natural language processing methods, can prove to be ineffective when dealing with the very large numbers of hypertext documents of different structure and subject. Moreover, such systems tend to treat HTML documents as plain texts not taking into account the additional information contained in their markup tags.