Results 1  10
of
35
Robust Classification for Imprecise Environments
, 1989
"... In realworld environments it is usually difficult to specify target operating conditions precisely. This uncertainty makes building robust classification systems problematic. We present a method for the comparison of classifier performance that is robust to imprecise class distributions and misclas ..."
Abstract

Cited by 255 (14 self)
 Add to MetaCart
In realworld environments it is usually difficult to specify target operating conditions precisely. This uncertainty makes building robust classification systems problematic. We present a method for the comparison of classifier performance that is robust to imprecise class distributions and misclassification costs. The ROC convex hull method combines techniques from ROC analysis, decision analysis and computational geometry, and adapts them to the particulars of analyzing learned classifiers. The method is efficient and incremental, minimizes the management of classifier performance data, and allows for clear visual comparisons and sensitivity analyses. We then show that it is possible to build a hybrid classifier that will perform at least as well as the best available classifier for any target conditions. This robust performance extends across a wide variety of comparison frameworks, including the optimization of metrics such as accuracy, expected cost, lift, precision, recall, and ...
Flexible Metric Nearest Neighbor Classification
, 1994
"... The Knearestneighbor decision rule assigns an object of unknown class to the plurality class among the K labeled "training" objects that are closest to it. Closeness is usually defined in terms of a metric distance on the Euclidean space with the input measurement variables as axes. The metric cho ..."
Abstract

Cited by 123 (2 self)
 Add to MetaCart
The Knearestneighbor decision rule assigns an object of unknown class to the plurality class among the K labeled "training" objects that are closest to it. Closeness is usually defined in terms of a metric distance on the Euclidean space with the input measurement variables as axes. The metric chosen to define this distance can strongly effect performance. An optimal choice depends on the problem at hand as characterized by the respective class distributions on the input measurement space, and within a given problem, on the location of the unknown object in that space. In this paper new types of Knearestneighbor procedures are described that estimate the local relevance of each input variable, or their linear combinations, for each individual point to be classified. This information is then used to separately customize the metric used to define distance from that object in finding its nearest neighbors. These procedures are a hybrid between regular Knearestneighbor methods and treestructured recursive partitioning techniques popular in statistics and machine learning.
Logic regression
 Journal of Computational and Graphical Statistics
, 2003
"... Logic regression is an adaptive regression methodology that attempts to construct predictors as Boolean combinations of binary covariates. In many regression problems a model is developed that relates the main effects (the predictors or transformations thereof) to the response, while interactions ar ..."
Abstract

Cited by 36 (11 self)
 Add to MetaCart
Logic regression is an adaptive regression methodology that attempts to construct predictors as Boolean combinations of binary covariates. In many regression problems a model is developed that relates the main effects (the predictors or transformations thereof) to the response, while interactions are usually kept simple (two to threeway interactions at most). Often, especially when all predictors are binary, the interaction between many predictors may be what causes the differences in response. This issue arises, for example, in the analysis of SNP microarray data or in some data mining problems. In the proposed methodology, given a set of binary predictors we create new predictors such as “X1, X2, X3, and X4 are true, ” or “X5 or X6 but not X7 are true. ” In more speci � c terms: we try to � t regression models of the form g(E[Y]) = b0 + b1L1 + ¢ ¢ ¢ + bnLn, where Lj is any Boolean expression of the predictors. The Lj and bj are estimated simultaneously using a simulated annealing algorithm. This article discusses how to � t logic regression models, how to carry out model selection for these models, and gives some examples.
Statistical Learning Algorithms Based on Bregman Distances
, 1997
"... We present a class of statistical learning algorithms formulated in terms of minimizing Bregman distances, a family of generalized entropy measures associated with convex functions. The inductive learning scheme is akin to growing a decision tree, with the Bregman distance filling the role of the im ..."
Abstract

Cited by 23 (1 self)
 Add to MetaCart
We present a class of statistical learning algorithms formulated in terms of minimizing Bregman distances, a family of generalized entropy measures associated with convex functions. The inductive learning scheme is akin to growing a decision tree, with the Bregman distance filling the role of the impurity function in treebased classifiers. Our approach is based on two components. In the feature selection step, each linear constraint in a pool of candidate features is evaluated by the reduction in Bregman distance that would result from adding it to the model. In the constraint satisfaction step, all of the parameters are adjusted to minimize the Bregman distance subject to the chosen constraints. We introduce a new iterative estimation algorithm for carrying out both the feature selection and constraint satisfaction steps, and outline a proof of the convergence of these algorithms. 1 Introduction In this paper we present a class of statistical learning algorithms formulated in terms...
Condensing Uncertainty via Incremental Treatment Learning
 ANNALS OF SOFTWARE ENGINEERING, SPECIAL ISSUE ON COMPUTATIONAL INTELLIGENCE. TO APPEAR.
, 2002
"... Models constrain the range of possible behaviors de£ned for a domain. When parts of a model are uncertain, the possible behaviors may be a data cloud: i.e. an overwhelming range of possibilities that bewilder an analyst. Faced with large data clouds, it is hard to demonstrate that any particular de ..."
Abstract

Cited by 22 (18 self)
 Add to MetaCart
Models constrain the range of possible behaviors de£ned for a domain. When parts of a model are uncertain, the possible behaviors may be a data cloud: i.e. an overwhelming range of possibilities that bewilder an analyst. Faced with large data clouds, it is hard to demonstrate that any particular decision leads to a particular outcome. Even if we can’t make de£nite decisions from such models, it is possible to £nd decisions that reduce the variance of values within a data cloud. Also, it is possible to change the range of these future behaviors such that the cloud condenses to some improved mode. Our approach uses two tools. Firstly, a model simulator is constructed that knows the range of possible values for uncertain parameters. Secondly, the TAR2 treatment learner uses the output from the simulator to incrementally learn better constraints. In our incremental treatment learning cycle, users review newly discovered treatments before they are added to a growing pool of constraints used by the model simulator.
Handling missing values when applying classification models. Journal of machine learning research
, 2007
"... Much work has studied the effect of different treatments of missing values on model induction, but little work has analyzed treatments for the common case of missing values at prediction time. This paper first compares several different methods—predictive value imputation, the distributionbased impu ..."
Abstract

Cited by 16 (4 self)
 Add to MetaCart
Much work has studied the effect of different treatments of missing values on model induction, but little work has analyzed treatments for the common case of missing values at prediction time. This paper first compares several different methods—predictive value imputation, the distributionbased imputation used by C4.5, and using reduced models—for applying classification trees to instances with missing values (and also shows evidence that the results generalize to bagged trees and to logistic regression). The results show that for the two most popular treatments, each is preferable under different conditions. Strikingly the reducedmodels approach, seldom mentioned or used, consistently outperforms the other two methods, sometimes by a large margin. The lack of attention to reduced modeling may be due in part to its (perceived) expense in terms of computation or storage. Therefore, we then introduce and evaluate alternative, hybrid approaches that allow users to balance between more accurate but computationally expensive reduced modeling and the other, less accurate but less computationally expensive treatments. The results show that the hybrid methods can scale gracefully to the amount of investment in computation/storage, and that they outperform imputation even for small investments.
An Analysis of Reduced Error Pruning
 Journal of Artificial Intelligence Research
, 2001
"... Topdown induction of decision trees has been observed to suffer from the inadequate functioning of the pruning phase. In particular, it is known that the size of the resulting tree grows linearly with the sample size, even though the accuracy of the tree does not improve. Reduced Error Pruning is a ..."
Abstract

Cited by 13 (4 self)
 Add to MetaCart
Topdown induction of decision trees has been observed to suffer from the inadequate functioning of the pruning phase. In particular, it is known that the size of the resulting tree grows linearly with the sample size, even though the accuracy of the tree does not improve. Reduced Error Pruning is an algorithm that has been used as a representative technique in attempts to explain the problems of decision tree learning. In this paper we present analyses of Reduced Error Pruning in three different settings. First we study the basic algorithmic properties of the method, properties that hold independent of the input decision tree and pruning examples. Then we examine a situation that intuitively should lead to the subtree under consideration to be replaced by a leaf node, one in which the class label and attribute values of the pruning examples are independent of each other. This analysis is conducted under two different assumptions. The general analysis shows that the pruning probability of a node fitting pure noise is bounded by a function that decreases exponentially as the size of the tree grows. In a specific analysis we assume that the examples are distributed uniformly to the tree. This assumption lets us approximate the number of subtrees that are pruned because they do not receive any pruning examples. This paper clarifies the different variants of the Reduced Error Pruning algorithm, brings new insight to its algorithmic properties, analyses the algorithm with less imposed assumptions than before, and includes the previously overlooked empty subtrees to the analysis.
Using Bayesian neural networks to classify forest scenes
, 1998
"... We present results that compare the performance of Bayesian learning methods for neural networks on the task of classifying forest scenes into trees and background. Classification task is demanding due to the texture richness of the trees, occlusions of the forest scene objects and diverse lighting ..."
Abstract

Cited by 10 (10 self)
 Add to MetaCart
We present results that compare the performance of Bayesian learning methods for neural networks on the task of classifying forest scenes into trees and background. Classification task is demanding due to the texture richness of the trees, occlusions of the forest scene objects and diverse lighting conditions under operation. This makes it difficult to determine which are optimal image features for the classification. A natural way to proceed is to extract many different types of potentially suitable features, and to evaluate their usefulness in later processing stages. One approach to cope with large number of features is to use Bayesian methods to control the model complexity. Bayesian learning uses a prior on model parameters, combines this with evidence from a training data, and then integrates over the resulting posterior to make predictions. With this method, we can use large networks and many features without fear of overfitting. For this classification task we compare two Bayesian l...
Generalization And Discrimination In TreeStructured Unit Selection
 in Proceedings of the 3rd ESCA/COCOSDA International Speech Synthesis Workshop
, 1998
"... Concatenative "selectionbased" synthesis from large databases has emerged as a viable framework for TTS waveform generation. Unit selection algorithms attempt to predict the appropriateness of a particular database speech segment using only linguistic features output by text analysis and prosody pr ..."
Abstract

Cited by 8 (0 self)
 Add to MetaCart
Concatenative "selectionbased" synthesis from large databases has emerged as a viable framework for TTS waveform generation. Unit selection algorithms attempt to predict the appropriateness of a particular database speech segment using only linguistic features output by text analysis and prosody prediction components of a synthesizer. All of these algorithms have in common a training or "learning" phase in which parameters are trained to select appropriate waveform segments for a given feature vector input. One approach to this step is to partition available data into clusters that can be indexed by linguistic features available at runtime. This method relies critically on two important principles: discrimination of fine phonetic details using a perceptuallymotivated distance measure in training and generalization to unseen cases in selection. In this paper, we describe efforts to systematically investigate and improve these parts of the process.
Exemplar Learning in Fuzzy Decision Trees
 in Proceedings of FUZZIEEE
, 1996
"... Decisiontree algorithms provide one of the most popular methodologies for symbolic knowledge acquisition. The resulting knowledge, a symbolic decision tree along with a simple inference mechanism, has been praised for comprehensibility. The most comprehensible decision trees have been designed for ..."
Abstract

Cited by 7 (1 self)
 Add to MetaCart
Decisiontree algorithms provide one of the most popular methodologies for symbolic knowledge acquisition. The resulting knowledge, a symbolic decision tree along with a simple inference mechanism, has been praised for comprehensibility. The most comprehensible decision trees have been designed for perfect symbolic data. Over the years, additional methodologies have been investigated and proposed to deal with continuousor multivalueddata, and with missing or noisy features. Recently, with the growing popularity of fuzzy representation, a few researchers independently have proposed to utilize fuzzy representation in decision trees to deal with similar situations. Fuzzy representation bridges the gap between symbolic and nonsymbolic data by linking qualitatitive linguistic terms with quantitative data. In this paper, we overview our fuzzy decision tree and propose a few new inferences based on exemplar learning.