Optimized cutting plane algorithm for support vector machines

By Vojtech Franc, Sören Sonnenburg. 2008

133 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

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

1. Cordelia Schmid
2. Zeynep Akata
3. Bernt Schiele
4. Quoc V. Le
5. Mohit Bansal
6. Trevor Darrell
7. Andrew Zisserman
8. Yoshua Bengio

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Cordelia Schmid
3. Trevor Darrell
4. Bernt Schiele
5. Andrew Zisserman
6. Dacheng Tao
7. Andrea Vedaldi
8. Xiaogang Wang

Citations in the corpus

• Learning Optimized Risk Scores (2016, methodology): “ this experiment in Appendix B. Stalling Behavior on Non-Convex Problems Algorithm 1 is similar to cutting-plane algorithms used for large-scale machine learning problems (see e.g., Teo et al., 2009; Franc and Sonnenburg, 2008) in that it solves an optimization problem at each iteration (see Step 5)”
• Nonlinear optimization and support vector machines (2018, methodology): “…is defined for regularized risk minimization problems, that is shown to converge in O ( 1 /(cid:13)) steps for linear classification problems, and that is further optimized in Franc and Sonnenburg (2009) and Franc and Sonnenburg (2008), where an optimized choice of the cutting planes is described.”
• Bundle Methods for Regularized Risk Minimization (2010, methodology): “observed by Franc and Sonnenburg (2008) in the case of linear SVMs with binary hinge loss.”
• A Quasi-Newton Approach to Nonsmooth Convex Optimization Problems in Machine Learning (2008, methodology): “Therefore, subgradient-based approaches such as subgradient descent (Nedić and Bertsekas, 2000) or bundle methods (Joachims, 2006; Franc and Sonnenburg, 2008; Teo et al., 2009) have gained considerable attention for minimizing nonsmooth objectives.”
• Modelling input texts: from Tree Kernels to Deep Learning (2015, methodology): “Recently, a number of efficient methods to train SVMs based on the idea of the Cutting Plane Algorithm (CPA) have been proposed [46,69].”
• Good Practice in Large-Scale Learning for Image Classification (2014, methodology): “State-of-the art batch algorithms for linear SVMs are based on coordinate-descent approaches with second-order acceleration.”
• Recent Advances of Large-Scale Linear Classification (2012, methodology): “• L2-regularized L1-loss SVM: Available approaches include, for example, cutting plane methods for the primal form (SVM [4], OCAS [21], and BMRM [22]), a stochastic (sub-)gradient descent method for the primal form (Pegasos [5] and SGD [23]), and a coordinate descent method for the dual form (LIBLIN”
• Optimized Cutting Plane Algorithm for Large-Scale Risk Minimization (2009, methodology): “…Astro data sets from Joachims (2006) by the MNIST, an artificial dense data set and two larger bioinformatics splice site recognition data sets for worm and human.3 These data sets have been used and are described in detail in Joachims (2006), Shwartz et al. (2007) and Franc and Sonnenburg (2008a).”