Abstract not found

To Mom, Dad, and Susan, for their support and encouragement.

An important and difficult prediction task in many domains, particularly medical decision making, is that of prognosis. Prognosis presents a unique set of problems to a learning system when some of the outputs are unknown. This paper presents a new approach to prognostic prediction, using ideas from nonparametric statistics to fully utilize all of the available information in a neural architecture. The technique is applied to breast cancer prognosis, resulting in flexible, accurate models that may play a role in preventing unnecessary surgeries. 1 Introduction This paper applies artificial neural network classification to the analysis of survival or lifetime data (Lee, 1992), in which the objective can be broadly defined as predicting the future time of a particular event. In this work we are concerned specifically with prognosis, that is, predicting the course of a disease. These methods are applied to breast cancer prognosis, predicting how long after surgery we can expect the disea...

This paper presents an automatic segmentation method to locate cell nuclei in cytological images using two iterative generalized Hough Transforms (GHTs), one for obtaining knowledge on the size of the nuclei in the image, and the other for isolating nuclei themselves. In order to reduce the parameter space of the image, the original image is scaled down to half-sized and quarter-sized images. Using the information from the rst iterative GHT on a quarter-sized image, the range of nuclear sizes is determined to limit the parameter space of the half-sized image. After the second iterative GHT on the half-sized image, nuclei are detected and segmented with edge information which helps determine the exact boundary. The results show that this method gives reduction in computation time and memory usage without loss of accuracy. 1 Introduction Detecting objects in images is one of most interesting tasks of computer vision. An important example is cell detection in cytologic and hist...

A survey of several well known rule extraction techniques is presented in my report, in the light of a broader paradigm of connectionist-symbolic learning. In the first part of the report I have covered some introductory aspects about machine learning, and investigated the reasons for a possible connectionist-symbolic integration, thereby presenting a hybrid learning framework. Within this hybrid learning framework, my report focuses on the survey of Rule extraction techniques from Trained Artificial Neural Networks (ANNs). I have presented the techniques, and thereby compare them on several grounds, and investigate the relative merits and demerits of each one of them. Towards the end, there...

Although they are applicable to a wide array of problems, and have demonstrated good performance on a number of difficult, real-world tasks, neural networks are not usually applied to problems in which comprehensibility of the acquired concepts is important. The concept representations formed by neural networks are hard to understand because they typically involve distributed, nonlinear relationships encoded by a large number of real-valued parameters. To address this limitation, we have been developing algorithms for extracting "symbolic" concept representations from trained neural networks. We first discuss why it is important to be able to understand the concept representations formed by neural networks. We then briefly describe our approach and discuss a number of issues pertaining to comprehensibility that have arisen in our work. Finally, we discuss choices that we have made in our research to date, and open research issues that we have not yet addressed. 1 Introduction Neural ...

Feature selection plays an important role in pattern classification. In this paper, we present an improved Branch and Bound algorithm for optimal feature subset selection. This algorithm searches for an optimal solution in a large solution tree in an efficient manner by cutting unnecessary paths which are guaranteed not to contain the optimal solution. Our experimental results demonstrate the effectiveness of the new algorithm.

The generalized Hough transform (GHT) is useful for detecting and segmenting 2dimensional (2D) object shapes. However, a weakness of GHT is that one must enumerate many templates corresponding to variation of the desired shape. This paper presents a new approach that uses unsupervised learning to nd a set of templates specic to the objects being outlined by the user. The templates are formed by averaging the shapes that belong to a particular cluster. Further, uncertainty regions are added by calculating the standard deviation of points on the template of the cluster. The eectiveness of the resulting system is demonstrated on a medical diagnosis task using cytological images. Our approach can serve as a fully automated substitute to the tedious and time-consuming task of manual shape registration and analysis. Keywords: generalized Hough transform, exible templates, possibilities, segmentation, unsupervised learning 1 Introduction A main task of computer vision is to locate obje...

COMPUTATIONAL ENGINEERING AND DESIGN CENTER SCHOOL OF ENGINEERING SCIENCES Doctor of Philosophy Artificial Intelligence Technologies in Complex Engineering Design by Yew Soon Ong Engineering design optimization is an emerging technology whose application both tends to shorten design-cycle time and finds new designs that are not only feasible, but also nearer to optimum, based on specified design criteria. Its gain in attention in the field of complex designs is fuelled by advancing computing power now allowing increasingly accurate analysis codes to be deployed. Unfortunately, the optimization of complex engineering design problems remains a di#cult task, due to the complexity of the cost surfaces and the human expertise necessary in order to achieve high quality results. This research is concerned with the e#ective use of past experiences and chronicled data from previous designs to mitigate some of the limitations of present engineering design optimization process. In particular, the present work leverages well established artificial intelligence technologies and extends recent theoretical and empirical advances, particularly in machine learning, adaptive hybrid evolutionary computation, surrogate modeling, radial basis functions and transductive inference, to mitigate the issues of i) choice of optimization methods and ii) dealing with expensive design problems. The resulting approaches are studied using commonly employed benchmark functions. Further demonstrations on realistic aerodynamic aircraft and ship design problems reveal that the proposed techniques not only generate robust design performance, they can also greatly decrease the cost of design space search and arrive at better designs as compared to conventional approaches.

The prediction of survival time or recurrence time is an important learning problem in medical domains. The Recurrence Surface Approximation (RSA) method is a natural, effective method for predicting recurrence times using censored input data. This paper introduces the Survival Curve RSA (SC-RSA), an extension to the RSA approach which produces accurate predicted rates of recurrence, while maintaining accuracy on individual predicted recurrence times. The method is applied to the problem of breast cancer recurrence using two different datasets. 1 Introduction A common prediction problem in many different fields is the analysis of survival or lifetime data [13, 15], in which the objective can be broadly defined as predicting the time of a particular event. Specifically, in medicine, we are concerned with prognosis, that is, predicting the course of a disease based on known patient characteristics. The "event time" to be estimated could be either the death of the patient or the recurr...