A Practical Guide to Support Vector Classification

By Chih-Wei Hsu, Chih-Chung Chang, Chih-Jen Lin. 2009

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

AutoML: A Survey of the State-of-the-Art (2019, methodology): “[97] recommend using a coarse grid first, and after finding a better region on the gird, they implement a finer grid search on that region.”
Survey on Automated Machine Learning (2019, background): “Therefore, each continuous hyperparameter is discretized into k (logarithmic) equidistant values λ (i) 1:k; for categorical hyperparameters each value is used (Hsu et al., 2003).”
Knowledge Matters: Importance of Prior Information for Optimization (2013, background): “For the polynomial kernel, γ controls the ﬂexibility of the classiﬁer (degree of the polynomial) as the number of parameters increases (Hsu et al., 2003; Ben-Hur and Weston, 2010).”
The effect of machine learning regression algorithms and sample size on individualized behavioral prediction with functional connectivity features (2018, unknown): “, 29, 210] (Hsu et al., 2003), and accordingly, λ from among 16 values [2 10, 2-,9, .”
Alibaba Submission to the WMT18 Parallel Corpus Filtering Task (2018, methodology): “Here, the weights of these scores are selected with gird search method(Hsu et al., 2003).”
ProQ3: Improved model quality assessments using Rosetta energy terms (2016, background): “SVM predictor works best when the input features are either scaled between -1 and 1 or between 0 and 1 [22].”
A second-order method for strongly convex ℓ1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\ell _1$$\end{document}-re (2013, methodology): “The last column shows the τ that gave the classificationwith the highest accuracy after performing afivefold cross validationover various τ values, as proposed in [13].”
A Review of Non-Invasive Techniques to Detect and Predict Localised Muscle Fatigue (2011, methodology): “SVMs are a useful technique of data classification [220,221].”
Support vector machine in structural reliability analysis: A review (2023, unknown): “…techniques were developed to search the SVM hyperparameters e.g., gradient descent algorithm [32], coordinate descent method [33], grid search [34], five-fold cross-validation method [24], real-value genetic algorithm [35], particle swarm optimization [36], dynamic particle filter [37],…”
Autoreject: Automated artifact rejection for MEG and EEG data (2016, methodology): “Finally, the best parameters ρ ∗ and κ ∗ are estimated using grid search (Hsu et al., 2003).”
Code Mixing: A Challenge for Language Identification in the Language of Social Media (2014, methodology): “…(linear kernel) for different combination of these features.3 Parameter optimizations (C range 2-15 to 210) for SVM are performed for each feature 3According to (Hsu et al., 2010) the SVM linear kernel with parameter C optimization is good enough when dealing with a large number of features.”
Painful data: The UNBC-McMaster shoulder pain expression archive database (2011, methodology): “A linear kernel was used in our experiments due to its ability to generalize well to unseen data in many pattern recognition tasks [26].”
Trajectories of depression: unobtrusive monitoring of depressive states by means of smartphone mobility traces analysis (2015, methodology): “For each user, we train and test a personalized Support Vector Machine (SVM) classifier with a Gaussian radial basis function kernel [14].”
Decision Tree and SVM-Based Data Analytics for Theft Detection in Smart Grid (2016, methodology): “Then, grid-search method is used for finding the values of C and γ [36].”
Churn prediction in subscription services: An application of support vector machines while comparing two parameter-selection techniques (2008, methodology): “Basically, all combinations of (C, c) are tried and two pairs of parameters are restrained: (1) the one with the best crossvalidated accuracy – as proposed by Hsu et al., 2004 – and (2) the one with the biggest cross-validated area under the receiver operating curve.”