Description:
Many existing researches utilized many different approaches for recognition in digital mammography using various ANN classifier-modeling techniques. Different types of feature extraction techniques are also used. It has been observed that, beyond a certain point, the inclusion of additional features leads to a worse rather than better performance. Moreover, the choice of features to represent the patterns affects several aspects of pattern recognition problem such as accuracy, required learning time and necessary number of samples. A common problem with the multi category feature classification is the conflict between the categories. None of the feasible solutions allow simultaneous optimal solution for all categories. In order to find an optimal solutions the searching space can be divided based on individual category in each sub region and finally merging them through decision spport system. In this paper we propose a canonical GA based modular feature selection approach combined with standard MLP.
Description:
Many existing researches utilized many different approaches for recognition in digital mammography using various ANN classifier-modeling techniques. Different types of feature extraction techniques are also used. It has been observed that, beyond a certain point, the inclusion of additional features leads to a worse rather than better performance. Moreover, the choice of features to represent the patterns affects several aspects of pattern recognition problem such as accuracy, required learning time and necessary number of samples. A common problem with the multi category feature classification is the conflict between the categories. None of the feasible solutions allow simultaneous optimal solution for all categories. In order to find an optimal solutions the searching space can be divided based on individual category in each sub region and finally merging them through decision spport system. In this paper we propose a canonical GA based modular feature selection approach combined with standard MLP.
Description:
In this paper we incorporate a hybrid evolutionary method, which uses a combination of genetic algorithm and matrix based solution method such as QR factorization. A heuristic segmentation algorithm is initially used to over segment each word. Then the segmentation points are passed through the rule-based module to discard the incorrect segmentation points and include any missing segmentation points. Following the segmentation the connected contour is extracted between two correct segmentation points. The contour is passed through the feature extraction module that extracts the angular features of the contour, after which the EALS-BT algorithm finds the architecture and the weights for the classifier network. These recognized characters are grouped into words and passed to a variable length lexicon that retrieves words that has highest confidence value. Hamming neural network is used as a lexicon that rectifies the word misrecognized by the classifier. We have used CEDAR benchmark dataset and UCI Machine Learning repository (Upper case) to test the train and test the system
Description:
Modular neural networks have the possibility of overcoming common scalability and interference problems experienced by fully connected neural networks when applied to large databases. In this paper we trial an approach to constructing modular ANN's for a very large problem from CEDAR for the classification of handwritten characters. In our approach, we apply progressive task decomposition methods based upon clustering and regression techniques to find modules. We then test methods for combining the modules into ensembles and compare their structural characteristics and classification performance with that of an ANN having a fully connected topology. The results reveal improvements to classification rates as well as network topologies for this problem.
Description:
Modular neural networks have the possibility of overcoming common scalability and interference problems experienced by fully connected neural networks when applied to large databases. In this paper we trial an approach to constructing modular ANN's for a very large problem from CEDAR for the classification of handwritten characters. In our approach, we apply progressive task decomposition methods based upon clustering and regression techniques to find modules. We then test methods for combining the modules into ensembles and compare their structural characteristics and classification performance with that of an ANN having a fully connected topology. The results reveal improvements to classification rates as well as network topologies for this problem.
Description:
Extracts of three case studies, from a larger qualitative study, are used to illustrate an emerging theme about the issue of the risks versus the benefits of providing detailed information about adverse drug reactions (ADRs) to consumers. The case studies include the consumer, medical and expert views of a single event, a suspected adverse drug reaction. The consumers would like more information. The doctors expressed that providing information about all possible ADRs can result in some consumers choosing not to take medications due to the perceived risk of the medication. The ADR experts stated that providing ADR information to the consumers within the presented case studies, might have assisted in earlier detection of the ADR, resulting in a less severe reaction. This paper aims to illustrate this emerging theme of the risks versus the benefits of providing consumers with information about potential ADRs and suggest some ways of increasing the benefits and minimizing the risk. We then indicate how these finding may inform future development of ADR decision support systems.
Description:
Decision making processes in the health care setting are often complex, demanding of professionals and rapidly changing. Furthermore, there is increasing pressure for professionals to make reasoning transparent and consistent. Decision support technologies have not made a substantial contribution to these issues to date largely because knowledge is difficult to elicit and maintain and existing development tools are very sophisticated yet complex. In this study a method for representing complex and discretionary reasoning used successfully in law was applied to the task of modelling decision making processes that critical care nurses deploy in responding to a low oxygen alarm. The approach, based on decision and argument tree diagrams enables the rapid development of small scale, yet useful web based decision support systems.
Description:
The problem of cluster analysis is formulated as a problem of non-smooth, non-convex optimization, and an algorithm for solving the cluster analysis problem based on non-smooth optimization techniques is developed. We discuss applications of this algorithm in large databases. Results of numerical experiments are presented to demonstrate the effectiveness of this algorithm.