Working Set Selection Using Second Order Information for Training Support Vector Machines

By Rong-En Fan, Pai-Hsuen Chen, Chih-Jen Lin. Journal of machine learning research, 2005

1,561 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/faef8fb61fdc2ca13a34cfe9f991fd30b21bfe0f

How researchers see this paper

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

1. Yoshua Bengio
2. Dacheng Tao
3. Quoc V. Le
4. Yu Zhang
5. Trevor Darrell
6. Yi Yang
7. Chunhua Shen
8. Luc Van Gool

Citations in the corpus

• Efficient max-margin multi-label classification with applications to zero-shot learning (2012, methodology): “We can therefore use co-ordinate ascent algorithms (Platt 1999; Fan et al. 2005; Lin et al. 2009) to maximise the dual.”
• Support Vector Machine Solvers (2007, methodology): “The LIBSVM solver uses the following procedure (Fan et al., 2005). i = argmax k∈Iup ykgk j = argmax k∈Idown (yigi − ykgk) 2 2(Kii +Kkk − 2Kik) subject to yigi > ykgk (27) The computation of i is exactly as for the maximal violating pair scheme.”
• QP Algorithms with Guaranteed Accuracy and Run Time for Support Vector Machines (2006, methodology): “It is worth mentioning that other methods have been recntly introduced which examine a subset of pairs and choose the one with the largest stepwise improve ent (e.g. see Fan et al., 2005; Lai et al., 2003).”
• Hybrid support vector machine rule extraction method for discovering the preferences of stock market investors: Evidence from Montenegro (2015, methodology): “Fan, Chen, and Lin (2005), proposed SVM learner, which has been implemented using LIBSVM software since version 2.8 (Chang & Lin, 2011).”
• Person Identification Using Full-Body Motion and Anthropometric Biometrics from Kinect Videos (2012, background): “Since the computation complexity of the SVD and SVM algorithms are O(pq2 + p(2)q + q(3)) [21] and O(q2) [22] respectively, the computational complexity of the action classifier ≈ O(q3) where q = 2mn.”
• SOLpro: accurate sequence-based prediction of protein solubility (2009, methodology): “For kNN and NN we used Weka (Witten and Frank, 2005), and for SVM we used LIBSVM (Chang and Lin, 2001), which implements the Sequential Minimal Optimization (SMO) algorithm proposed in (Fan et al., 2005).”
• Parallel Multi-task Learning (2015, methodology): “In order to find the appropriate working set with linear complexity, similar to [28] we adopt a greedy strategy by setting j as j = argmaxt∈I1(αi) y i t▽h(αi)t, and choosing k ∈ I2(αi) such that (j, k) is a violating pair and the reduction on the objective function value of problem (14) after the up”
• Fast Stochastic Frank-Wolfe Algorithms for Nonlinear SVMs (2010, methodology): “• SMO-wss2: SMO with shrinking technique and second order information for working set selection [10].”
• Dual coordinate descent methods for logistic regression and maximum entropy models (2011, background): “Alternatively, using gradient information (e.g., Joachims, 1998; Keerthi et al., 2001; Fan et al., 2005) may lead to faster convergence.”
• Online chemical modeling environment (OCHEM): web platform for data storage, model development and publishing of chemical information (2011, methodology): “Support Vector Machine (SVM) uses the LibSVM program [78].”
• Wireless device identification with radiometric signatures (2008, methodology): “We implemented two rather different radiometric classifiers, one using support vector machines algorithm (SVM) [7, 9, 11] and the other using the k-nearest-neighbor (kNN) [24] algorithm.”