Feature selection based on mutual information criteria of max-dependency, max-relevance, and min-redundancy

By Hanchuan Peng, Fuhui Long, Chris H. Q. Ding. 2003

9,860 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Molecular Biology

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Molecular Biology

Citations in the corpus

Adversarial Feature Selection Against Evasion Attacks (2016, background): “the dataset [53]–[55], such as mutual information [56], [57].”
A practical tutorial on autoencoders for nonlinear feature fusion: Taxonomy, models, software and guidelines (2018, background): “The former approach usually relies on correlation or mutual information between input and output variables [24, 25], while the latter tends to avoid redundancy among features [26].”
Anti-Retroactive Interference for Lifelong Learning (2022, methodology): “In this way, model information redundancy can be avoided, which is consistent with the machine learning theory in [19].”
Mutual Information Neural Estimation (2018, background): “, 2000), feature selection (Kwak & Choi, 2002; Peng et al., 2005), and causality (Butte & Kohane, 2000).”
Learning Classification Models of Cognitive Conditions from Subtle Behaviors in the Digital Clock Drawing Test (2015, methodology): “From among all of the features, we created a subset of the first 200 selected by Minimum- Redundancy–Maximum-Relevance filtering (Peng et al. 2005).”
Information-Theoretic Feature Selection in Microarray Data Using Variable Complementarity (2008, methodology): “The SynTRen simulator is used to generate 14 different classification tasks where the target is the simulated expression of a TABLE X TABLE REPORTING THE NUMBER OF RELEVANT VARIABLES SELECTED BY MRMR AND MASSIVE ON 14 DATASETS OF 507 INPUT VARIABLES AND 100 (COLUMNS 2,3), 1000 (COLUMNS 4,5) AND 2000”
Variable selection in functional data classification: a maxima-hunting proposal (2013, methodology): “) = 1 card(S) X t∈S I(Xt,Y), Red(S) = 1 card2(S) X s,t∈S I(Xt,Xs), where card(S) denotes the cardinality of S. Then, the mRMR variables selection method, as proposed in Ding and Peng (2005) [see also Peng et al. (2005)], aims at maximizing the relevance avoiding an excess of redundancy. The algorith”
An unsupervised feature selection algorithm based on ant colony optimization (2014, methodology): “According to this fact, many studies have shown that removing redundant features can improve the classifier performance (Biesiada and Duch, 2007; Gu et al., 2011; Martínez Sotoca and Pla, 2010; Peng et al., 2005); (3) the filter approach result is based on only one iteration process, and the optimal”
Opcode sequences as representation of executables for data-mining-based unknown malware detection (2013, methodology): “We used Mutual Information [53] (shown in equation 1) to measure the statistical dependence of the two variables: where X is the opcode frequency and Y is the class of the ﬁle (i.e., malware or benign software), p ( x, y ) is the joint probability distribution function of X and Y , and p ( x ) and…”
A Hybrid Intelligent System Framework for the Prediction of Heart Disease Using Machine Learning Algorithms (2018, background): “0e mRMR does not take care of the joint association of features [24].”
mr2PSO: A maximum relevance minimum redundancy feature selection method based on swarm intelligence for support vector machine classification (2011, background): “In the classification phase, a point x is assigned a label y according to y = sgn[w x + b].”
SVM-RFE With MRMR Filter for Gene Selection (2010, methodology): “The effectiveness of this ratio is reﬂected by its consistently good performance shown on various expression datasets than by other criterion, such as the difference between relevancy and redundancy [6], [7].”
Scale Space and Variational Methods in Computer Vision (2015, methodology): “More recent approaches, such as [12,16] for colour images or [22] for symmetric matrices, rely on the same minimisation but without the restriction to the given data points.”