Feature Selection

By Jundong Li, Kewei Cheng, Suhang Wang, Fred Morstatter, Robert P. Trevino, Jiliang Tang, Huan Liu. 2019

1,437 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Statistics and Probability

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 Statistics and Probability

Citations in the corpus

AutoML: A Survey of the State-of-the-Art (2019, background): “Feature extraction and construction are variants of feature transformation, by which a new set of features is created [82].”
MorphNet: Fast & Simple Resource-Constrained Structure Learning of Deep Networks (2017, methodology): “Later work [32] proposed to use an L1 regularization, which is known to induce sparsity [28, 24], thus alleviating the need for sophisticated estimates of a parameter’s contribution to the loss.”
Causal Interpretability for Machine Learning - Problems, Methods and Evaluation (2020, methodology): “ult to interpret the eﬀect of each feature. Fortunately, there are several methods that can be used to ﬁnd the importance of a feature in a linear regression such as t-statistics and chi-square score [57]. The aforementioned methods are restricted by users’ limitations (i.e., human understanding). W”
Tree-Guided Group Lasso for Multi-Task Regression with Structured Sparsity (2009, positive): “, a gene expression measure for a single gene), techniques such as lasso [10] or L1-regularized logistic regression [6, 11] have been developed to find a sparse and consistent regression function that identifies a parsimonious subset of inputs that determine the outputs.”
PHOTOMETRIC SUPERNOVA CLASSIFICATION WITH MACHINE LEARNING (2016, methodology): “While there are many other methods to determine the relevance of each feature (see Li et al. (2016) for a summary), we only use this simple measure of feature relevance to gain insight about the features and use the full classification pipeline to evaluate the effectiveness of our chosen features,…”
Data Classification: Algorithms and Applications (2014, background): “A number of implementations of neural network methods have been studied in [35,57,66,77,88], and many of these implementations are designed in the context of text data.”
Model-based clustering of high-dimensional data: A review (2014, neutral): “We refer to [80, 81] for further details on the effect of the dimensionality on classification in the asymptotic framework, i.”
Machine learning in human movement biomechanics: Best practices, common pitfalls, and new opportunities (2018, positive): “A lack of quantitative methods to assess motor symptoms associated with Parkinson’s disease has motivated the development of predictive models for automated detection of Parkinsonian gait, bradikynesia, tremor, freezing of gait, and medication-induced dyskinesia using data from accelerometers (Palme”
Probabilistic Multi-Label Classification with Sparse Feature Learning (2013, positive): “…logistic regression training to learn sparse label dependency features: min θ k − N i=1 log p(y i k |x i , y i k , θ k )+ (3) λ 1 2 w k 2 + λ 2 2 v k 2 + γv k 1 As one of the simplest sparsity inducing norm, 1 -norm has been widely used for feature selection [Ng, 2004; Zou and Hastie, 2005].”
Positive approximation: An accelerator for attribute reduction in rough set theory (2010, background): “Therefore, the omission of some attributes could not only be tolerable but even desirable relatively to the costs involved in such cases [32].”
Trading Accuracy for Sparsity in Optimization Problems with Sparsity Constraints (2010, positive): “Similarly, [20] considered PAC (probably approximately correct) learning of a sparse predictor and showed that 1 norm regularization is competitive with the best sparse predictor, while 2-regularization does not appear to be.”
Fast inference of deep neural networks in FPGAs for particle physics (2018, background): “L1 regularization is known to produce sparse models, provide built-in feature selection [81], and is a readily available option in many machine learning workflows.”
Selecting appropriate machine learning methods for digital soil mapping (2020, methodology): “Some popular variable selection methods include weight decay [73], LASSO (least absolute shrinkage”