http://hdl.handle.net/2262/2538 Tue, 21 May 2013 10:54:30 GMT 2013-05-21T10:54:30Z http://hdl.handle.net/2262/19021 Title: On the use of CBR in optimisation problems such as the TSP Author: Cunningham, Pádraig; Smyth, Barry; Hurley, Neil Abstract: The particular strength of CBR is normally considered to be its use in weak theory domains where solution quality is compiled into cases and is reusable. In this paper we explore an alternative use of CBR in optimisation problems where cases represent highly optimised structures in a huge highly constrained solution space. Our analysis focuses on the Travelling Salesman Problem where difficulty arises from the computational complexity of the problem rather than any difficulty associated with the domain theory. We find that CBR is good for producing medium quality solutions in very quick time. We have difficulty getting CBR to produce high quality solutions because solution quality seems to be lost in the adaptation process. We also argue that experiments with CBR on transparent problems such as the TSP tell us a lot about aspects of CBR such as; the quality of CBR solutions, the coverage that cases in the casebase offer and the utility of extending a case-base. Wed, 31 May 1995 23:00:00 GMT http://hdl.handle.net/2262/19021 1995-05-31T23:00:00Z http://hdl.handle.net/2262/16520 Title: Wireless Communication Using Real-Time Extensions to the Linux Network Subsystem Author: CAHILL, VINNY Abstract: Timely wireless communication is essential to allow real-time mobile applications, e.g., communication between mobile robots and intervehicle communication to be realized. The current IEEE 802.11 ad hoc protocol is unable to provide real-time communication guarantees due to its underlying contention-based MAC layer. Our current research is addressing the implementation of a time-bounded MAC protocol as a layer above 802.11. The implementation of a timely MAC protocol requires predictable and deterministic behavior at the device driver level, currently unavailable in the Linux operating system. This paper describes real-time extensions to the Linux operating system to provide real-time guarantees at the device driver level. To our knowledge, we are the first to implement a real-time ORiNOCO driver for real-time Linux. In addition we provide a low-level evaluation of the timeliness of packet transmission achievable using IEEE 802.11. Description: PUBLISHED Sun, 01 Jan 2006 00:00:00 GMT http://hdl.handle.net/2262/16520 2006-01-01T00:00:00Z http://hdl.handle.net/2262/16519 Title: Real-Time Communication in IEEE 802.11 Mobile Ad hoc Networks A Feasibility Study Author: CAHILL, VINNY; WEBER, STEFAN; GLEESON, MARK Abstract: Achieving predictable communication latency in an ad hoc IEEE 802.11 wireless local area network necessitates an approach that overcomes the impact of the underlying non-deterministic contention-based medium access control (MAC) protocol. In this paper, we assess the feasibility of using a Time Division Multiple Access (TDMA) layer above the IEEE 802.11 MAC protocol in order to achieve such predictable communication latency. We present the design and implementation of this TDMA layer and describe the influence of both the characteristics of the IEEE 802.11MAC protocol and of the dynamics of an ad hoc network on its design. We show that our approach can yield the predictability and stability required to support real-time communication in a real ad hoc environment, typified by dynamic host mobility and varied offered load, subject to identified constraints. Description: PUBLISHED Sun, 01 Jan 2006 00:00:00 GMT http://hdl.handle.net/2262/16519 2006-01-01T00:00:00Z http://hdl.handle.net/2262/13518 Title: Practical Solutions to the Problem of Diagonal Dominance in Kernel Document Clustering Author: Greene, Derek; Cunningham, Pádraig Abstract: In supervised kernel methods, it has been observed that the performance of the SVM classifier is poor in cases where the diagonal entries of the Gram matrix are large relative to the off-diagonal entries. This problem, referred to as diagonal dominance, often occurs when certain kernel functions are applied to sparse high-dimensional data, such as text corpora. In this paper we investigate the implications of diagonal dominance for unsupervised kernel methods, specifically in the task of document clustering. We discuss a selection of strategies for addressing this issue, and evaluate their effectiveness in producing more accurate and stable clusterings. Tue, 07 Feb 2006 00:00:00 GMT http://hdl.handle.net/2262/13518 2006-02-07T00:00:00Z http://hdl.handle.net/2262/13506 Title: ECUE: A Spam Filter that Uses Machine Learning to Track Concept Drift Author: Delany, Sarah Jane; Cunningham, Pádraig Abstract: While text classification has been identified for some time as a promising application area for Artificial Intelligence, so far few deployed applications have been described. In this paper we present a spam filtering system that uses example-based machine learning techniques to train a classifier from examples of spam and legitimate email. This approach has the advantage that it can personalise to the specifics of the user’s filtering preferences. This classifier can also automatically adjust over time to account for the changing nature of spam (and indeed changes in the profile of legitimate email). A significant software engineering challenge in developing this system was to ensure that it could interoperate with existing email systems to allow easy managment of the training data over time. This system has been deployed and evaluated over an extended period and the results of this evaluation are presented here. Fri, 10 Feb 2006 00:00:00 GMT http://hdl.handle.net/2262/13506 2006-02-10T00:00:00Z http://hdl.handle.net/2262/13505 Title: Predicting Probability Distributions for Surf Height Using an Ensemble of Mixture Density Networks Author: Carney, Michael; Cunningham, Pádraig; Dowling, Jim Abstract: There is a range of potential applications of Machine Learning where it would be more useful to predict the probability distribution for a variable rather than simply the most likely value for that variable. In meteorology and in finance it is often important to know the probability of a variable falling within (or outside) different ranges. In this paper we consider the prediction of surf height with the objective of predicting if it will fall within a given ‘surfable’ range. Prediction problems such as this are considerably more difficult if the distribution of the phenomenon is significantly different from a normal distribution. This is the case with the surf data we have studied. To address this we use an ensemble of mixture density networks to predict the probability density function. Our evaluation shows that this is an effective solution. We also describe a web-based application that presents these predictions in a usable manner. Fri, 10 Feb 2006 00:00:00 GMT http://hdl.handle.net/2262/13505 2006-02-10T00:00:00Z http://hdl.handle.net/2262/13504 Title: Calibrating Probability Density Forecasts with Multi-objective Search Author: Carney, Michael; Cunningham, Pádraig Abstract: In this paper, we show that the optimization of density forecasting models for regression in machine learning can be formulated as a multi-objective problem.We describe the two objectives of sharpness and calibration and suggest suitable scoring metrics for both.We use the popular negative log-likelihood as a measure of sharpness and the probability integral transform as a measure of calibration. We show how optimization on negative log-likelihood alone often results in sub-optimal models. To solve this problem we introduce a multi-objective evolutionary optimization framework that can produce better density forecasts from a prediction users perspective. Our experiments show improvements over state-of-the-art approaches. Fri, 10 Feb 2006 00:00:00 GMT http://hdl.handle.net/2262/13504 2006-02-10T00:00:00Z http://hdl.handle.net/2262/13503 Title: An Evaluation of the Usefulness of Explanation in a CBR System for Decision Support in Bronchiolitis Treatment Author: Doyle, Dónal; Cunningham, Pádraig Abstract: The research presented here explores the hypothesis that the deployment and acceptance of decision support systems in medicine will be enhanced if the basis for the recommendation produced by the system is apparent. We describe a decision support system for advising on patients suffering from bronchiolitis. This system supports its recommendations with precedent cases selected to support the recommendation along with justification text that highlights aspects of these cases relevant to the query case. It also presents an estimate of its confidence in the recommendation. The main contribution of this paper is an evaluation of this system in a clinical context. The evaluation shows that this type of explanation does enhance the usefulness of the system for practitioners. Mon, 03 Apr 2006 23:00:00 GMT http://hdl.handle.net/2262/13503 2006-04-03T23:00:00Z http://hdl.handle.net/2262/13502 Title: Evaluating Density Forecasting Models Author: Carney, Michael; Cunningham, Pádraig Abstract: Density forecasting in regression is gaining popularity as real world applications demand an estimate of the level of uncertainty in predictions. In this paper we describe the two goals of density forecasting1 sharpness and calibration. We review the evaluation methods available to a density forecaster to assess each of these goals and we introduce a new evaluation method that allows modelers to compare and evaluate their models across both of these goals simultaneously and identify the optimal model. Mon, 01 May 2006 23:00:00 GMT http://hdl.handle.net/2262/13502 2006-05-01T23:00:00Z http://hdl.handle.net/2262/13501 Title: Efficient Prediction-Based Validation for Document Clustering Author: Greene, Derek; Cunningham, Pádraig Abstract: Recently, stability-based techniques have emerged as a very promising solution to the problem of cluster validation. An inherent drawback of these approaches is the computational cost of generating and assessing multiple clusterings of the data. In this paper we present an efficient prediction-based validation approach suitable for application to large, high-dimensional datasets such as text corpora. We use kernel clustering to isolate the validation procedure from the original data. Furthermore, we employ a prototype reduction strategy that allows us to work on a reduced kernel matrix, leading to significant computational savings. To ensure that this condensed representation accurately reflects the cluster structures in the data, we propose a density-biased selection strategy. This novel validation process is evaluated on a large number of real and artificial datasets, where it is shown to consistently produce good estimates for the optimal number of clusters. Mon, 01 May 2006 23:00:00 GMT http://hdl.handle.net/2262/13501 2006-05-01T23:00:00Z