Selection of Relevant Features and Examples in Machine Learning

By Avrim Blum, Pat Langley. 1997

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Character Recognition Systems: A Guide for Students and Practitioners (2007, methodology): “We merge the taxonomies proposed by [12, 77], resulting in a division of the feature selection into four categories: embedded, filter, wrapper, and hybrid, as shown in Figure 3.”
Correlation and variable importance in random forests (2013, methodology): “2004; Díaz-Uriarte and Alvarez de Andrés 2006; Louw and Steel 2006; Genuer et al. (2010)).”
Genetic programming for simultaneous feature selection and classifier design (2006, background): “learning algorithms. In this case, the relevant features are filtered out from the irrelevant ones prior to the learning. Good surveys on FS include [1], [ 10 ].”
Feature Selection: A literature Review (2014, background): “Definition 2: Strongly Relevant to the Samples/ Distribution [18] “A feature is strongly relevant to the sample if there exists a pair of examples that only differ in their assignment to and .”
Relative-Error CUR Matrix Decompositions (2007, background): ““Feature selection” is a broad area that addresses the choice of columns explicitly for dimension reduction, but the metrics there are typically optimization-based [14] or machine-learning based [10].”
Opcode sequences as representation of executables for data-mining-based unknown malware detection (2013, methodology): “Because both IS and FS are very eﬀective at reducing the size of the training set and helping to ﬁltrate and clean noisy data, thereby improving the accuracy of machine-learning classiﬁers [6, 21], we strongly encourage the use of these methods.”
Random Forest classification of Mediterranean land cover using multi-seasonal imagery and multi-seasonal texture (2012, background): “Feature selection (FS), is an approach for selecting a subset of relevant features for building robust learning models (Blum and Langley, 1997; Guyon and Elisseeff, 2003; Saeys et al., 2007).”
Single Subject Prediction of Brain Disorders in Neuroimaging: Promises and Pitfalls (2016, methodology): “Instead of feature selection based on univariate group-level statistical tests, more common filtering and wrapper methods should be used (Blum and Langley, 1997;Hall and Smith, 1998; Kohavi and John, 1997).”
Studying Aesthetics in Photographic Images Using a Computational Approach (2006, background): “Feature selection for classification purposes is a well-studied topic [5], with some recent work related specifically to feature selection for SVMs.”
A review of instance selection methods (2010, methodology): “Another way for finding sub optimal solutions in search problems is the sequential search which has been used in selection problems such as feature selection (Pudil et al. 1994; Blum and Langley 1997) and for instance selection (Olvera-López et al.”
Conditional random fields for activity recognition (2007, methodology): “Filter methods suffer in the presence of many weakly relevant features [10], which makes them poorly suited for activity recognition domains.”
A stability index for feature selection (2007, background): “Feature selection has been a focus of pattern recognition for a long time now [13, 4, 9, 2, 3, 7, 12, 10, 11].”
Simultaneous feature selection and clustering using mixture models (2004, background): “Clustering, or cluster analysis, is prevalent in any discipline that involves analysis of multivariate data.”
Forecasting the short-term metro passenger flow with empirical mode decomposition and neural networks (2012, background): “Generally, the feature selection approaches can be divided into two major categories: the filter approach (Blum and Langley, 1997) and the wrapper approach (Kohavi and John, 1997).”
On the Relationship Between Feature Selection and Classification Accuracy (2008, background): “The main idea of feature subset selection is to remove redundant or irrelevant features from the data set as they can lead to a reduction of the classification accuracy or clustering quality and to an unnecessary increase of computational cost (Blum and Langley, 1997), (Koller and Sahami, 1996).”