294 research outputs found
Improving k-nearest neighbour classification with distance functions based on receiver operating characteristics
No full textThe k-nearest neighbour (k-NN) technique, due to its interpretable nature, is a simple and very intuitively appealing method to address classification problems. However, choosing an appropriate distance function for k-NN can be challenging and an inferior choice can make the classifier highly vulnerable to noise in the data. In this paper, we propose a new method for determining a good distance function for k-NN. Our method is based on consideration of the area under the Receiver Operating Characteristics (ROC) curve, which is a well known method to measure the quality of binary classifiers. It computes weights for the distance function, based on ROC properties within an appropriate neighbourhood for the instances whose distance is being computed. We experimentally compare the effect of our scheme with a number of other well-known k-NN distance metrics, as well as with a range of different classifiers. Experiments show that our method can substantially boost the classification performance of the k-NN algorithm. Furthermore, in a number of cases our technique is even able to deliver better accuracy than state-of-the-art non k-NN classifiers, such as support vector machines
Enhanced services for targeted information retrieval by event extraction and data mining
No full textWhere Information Retrieval (IR) and Text Categorization delivers a set of (ranked) documents according to a query, users of large document collections would rather like to receive answers. Question-answering from text has already been the goal of the Message Understanding Conferences. Since then, the task of text understanding has been reduced to several more tractable tasks, most prominently Named Entity Recognition (NER) and Relation Extraction. Now, pieces can be put together to form enhanced services added on an IR system. In this paper, we present a framework which combines standard IR with machine learning and (pre-)processing for NER in order to extract events from a large document collection. Some questions can already be answered by particular events. Other questions require an analysis of a set of events. Hence, the extracted events become input to another machine learning process which delivers the final output to the user's question. Our case study is the public collection of minutes of plenary sessions of the German parliament and of petitions to the German parliament. --
On the automated extraction of regression knowledge from databases
No full textThe advent of inexpensive, powerful computing systems, together with the increasing amount of available data, conforms one of the greatest challenges for next-century information science. Since it is apparent that much future analysis will be done automatically, a good deal of attention has been paid recently to the implementation of ideas and/or the adaptation of systems originally developed in machine learning and other computer science areas. This interest seems to stem from both the suspicion that traditional techniques are not well-suited for large-scale automation and the success of new algorithmic concepts in difficult optimization problems. In this paper, I discuss a number of issues concerning the automated extraction of regression knowledge from databases. By regression knowledge is meant quantitative knowledge about the relationship between a vector of predictors or independent variables (x) and a scalar response or dependent variable (y). A number of difficulties found in some well-known tools are pointed out, and a flexible framework avoiding many such difficulties is described and advocated. Basic features of a new tool pursuing this direction are reviewed
Multiagent Reinforcement Learning for Urban Traffic Control Using Coordination Graphs
No full textpeer reviewedSince traffic jams are ubiquitous in the modern world, optimizing, the behavior of traffic lights for efficient traffic flow is a critically important goal. Though most current traffic lights use simple heuristic protocols, more efficient controllers can be discovered automatically via multiagent reinforcement learning where each agent controls a single traffic light. However, in previous work on this approach, agents select only locally optimal actions without coordinating their behavior. This paper extends this approach to include explicit coordination between neighboring traffic lights. Coordination is achieved using the max-plus algorithm, which estimates the optimal joint action by sending locally optimized messages among connected agents. This paper presents the first application of max-plus to a large-scale problem and thus verifies its efficacy in realistic settings. It also provides empirical evidence that max-plus performs well on cyclic graphs, though it has been proven to converge only for tree-structured graphs. Furthermore, it provides a new understanding of the properties a traffic network must have for such coordination to be beneficial and shows that max-plus outperforms previous methods on networks that possess those properties
Large-Scale Clustering through Functional Embedding
No full textWe present a new framework for large-scale data clustering. The main
idea is to modify functional dimensionality reduction techniques
to directly optimize over discrete labels using stochastic gradient
descent. Compared to methods like spectral clustering our approach
solves a single optimization problem, rather than an ad-hoc two-stage
optimization approach, does not require a matrix inversion, can easily
encode prior knowledge in the set of implementable functions, and
does not have an ?out-of-sample? problem. Experimental results on
both artificial and real-world datasets show the usefulness of our
approach
Integration issue in knowledge acquisition systems
No full textS.124-131108Special issue on knowledge acquisitio
- …
