The Infona portal uses cookies, i.e. strings of text saved by a browser on the user's device. The portal can access those files and use them to remember the user's data, such as their chosen settings (screen view, interface language, etc.), or their login data. By using the Infona portal the user accepts automatic saving and using this information for portal operation purposes. More information on the subject can be found in the Privacy Policy and Terms of Service. By closing this window the user confirms that they have read the information on cookie usage, and they accept the privacy policy and the way cookies are used by the portal. You can change the cookie settings in your browser.
Support vector machines introduced three important innovations to machine learning research: (a) the application of mathematical programming algorithms to solve optimization problems in machine learning, (b) the control of overfitting by maximizing the margin, and (c) the use of Mercer kernels to convert linear separators into non-linear decision boundaries in implicit spaces. Despite their attractiveness...
Data mining research has approached the problems of analyzing large data sets in two ways. Simplifying a lot, the approaches can be characterized as follows. The database approach has concentrated on figuring out what types of summaries can be computed fast, and then finding ways of using those summaries. The model-based approach has focused on first finding useful model classes and then fast ways...
Many graphics are used for decoration rather than for conveying information. Some purport to display information, but provide insufficient supporting evidence. Others are so laden with information that it is hard to see either the wood or the trees. Analysing large data sets is difficult and requires technically efficient procedures and statistically sound methods to generate informative visualisations...
This paper reports on a long-term inter-disciplinary research project that aims at analysing the complex phenomenon of expressive music performance with machine learning and data mining methods. The goals and general research framework of the project are briefly explained, and then a number of challenges to machine learning (and also to computational music analysis) are discussed that arise from the...
The focus on scalability to very large datasets has been a distinguishing feature of the KDD endeavour right from the start of the area. In the present stage of its development, the field has begun to seriously approach the issue, and a number of different techniques for scaling up KDD algorithms have emerged. Traditionally, such techniques are concentrating on the search aspects of the problem, employing...
This paper presents a missing link between Plotkin’s least general generalization formalism and generalization on the Order Sorted Feature (OSF) foundation. A feature term (or Ψ-term) is an extended logic term based on ordered sorts and is a normal form of an OSF-term. An axiomatic definition of Ψ-term generalization is given as a set of OSF clause generalization rules and the least generality of...
Reasoning and learning from cases are based on the concept of similarity often estimated by a distance. This paper presents LID, a learning technique adequate for domains where cases are best represented by relations among entities. LID is able to 1) define a similitude term, a symbolic description of what is shared between a problem and precedent cases; and 2) assess the importance of the relations...
This paper investigates the use of meta-learning to estimate the predictive accuracy of a classifier. We present a scenario where meta-learning is seen as a regression task and consider its potential in connection with three strategies of dataset characterization. We show that it is possible to estimate classifier performance with a high degree of confidence and gain knowledge about the classifier...
The aim of behavioural cloning is to synthesize artificial controllers that are robust and comprehensible to human understanding. To attain the two objectives we propose the use of the Incremental Correction model that is based on a closed-loop control strategy to model the reactive aspects of human control skills. We have investigated the use of three different representations to encode the artificial...
This paper proposes a generic extension to propositional rule learners to handle multiple-instance data. In a multiple-instance representation, each learning example is represented by a bag” of fixed-length feature vectors”.Such a representation,lying somewhere between propositional and first-order representation, offers a tradeoff between the two. Naive-RipperMi is one implementation of this extension...
Modern agent and mediator systems communicate to a multitude of Web information providers to better satisfy user requests. They use wrappers to extract relevant information from HTML responses and to annotate it with userdefined labels. A number of approaches exploit the methods of machine learning to induce instances of certain wrapper classes, by assuming the tabular structure of HTML responses...
The reinforcement learning algorithm TD(λ) applied to Markov decision processes is known to need added exploration in many cases. With the usual implementations of exploration in TD-learning, the feedback signals are either distorted or discarded, so that the exploration hurts the algorithm’s learning. The present article gives a modification of the TD-learning algorithm that allows exploration without...
We point out that value-based reinforcement learning, such as TDand Q-learning, is not applicable to games of imperfect information. We give a reinforcement learning algorithm for two-player poker based on gradient search in the agents’ parameter spaces. The two competing agents experiment with different strategies, and simultaneously shift their probability distributions towards more successful actions...
Relational reinforcement learning (RRL) is a learning technique that combines standard reinforcement learning with inductive logic programming to enable the learning system to exploit structural knowledge about the application domain. This paper discusses an improvement of the original RRL. We introduce a fully incremental first order decision tree learning algorithm TG and integrate this algorithm...
In simulation studies boosting algorithms seem to be susceptible to noise. This article applies Ada.Boost.M2 used with decision stumps to the digit recognition example, a simulated data set with attribute noise. Although the final model is both simple and complex enough, boosting fails to reach the Bayes error. A detailed analysis shows some characteristics of the boosting trials which influence the...
This paper studies the Iterative Double Clustering (IDC) meta-clustering algorithm, a new extension of the recent Double Clustering (DC) method of Slonim and Tishby that exhibited impressive performance on text categorization tasks [1]. Using synthetically gener ated data we empirically demonstrate that whenever the DC procedure is successful in recovering some of the structure hidden in the data,...
Branching programs are a generalization of decision trees. From the viewpoint of boosting theory the former appear to be exponentially more efficient. However, earlier experience demonstrates that such results do not necessarily translate to practical success. In this paper we develop a practical version of Mansour and McAllester’s [13] algorithm for branching program boosting. We test the algorithm...
Machine learning methods for classification problems commonly assume that the class values are unordered. However, in many practical applications the class values do exhibit a natural order—for example, when learning how to grade. The standard approach to ordinal classification converts the class value into a numeric quantity and applies a regression learner to the transformed data, translating the...
This paper investigates neural network training as a potential source of problems for benchmarking continuous, heuristic optimization algorithms. Through the use of a student-teacher learning paradigm, the error surfaces of several neural networks are examined using so-called fitness distance correlation, which has previously been applied to discrete, combinatorial optimization problems. The results...
This paper proposes a new algorithm for pattern extraction from Stratified Ordered Trees (SOT). It first describes the SOT data structure that renders possible a representation of structured sequential data. Then it shows how it is possible to extract clusters of similar recurrent patterns from any SOT. The similarity on which our clustering algorithm is based is a generalized edit distance, also...
Set the date range to filter the displayed results. You can set a starting date, ending date or both. You can enter the dates manually or choose them from the calendar.