Solving large-scale multiclass learning problems via an efficient support vector classifier

Support vector machines （SVMs） are initially designed for binary classification. How to effectively extend them for multiclass classification is still an ongoing research topic. A multiclass classifier is constructed by combining SVM^light algorithm with directed acyclic graph SVM （DAGSVM） method, named DAGSVM^light A new method is proposed to select the working set which is identical to the working set selected by SVM^light approach. Experimental results indicate DAGSVM^light is competitive with DAGSMO. It is more suitable for practice use. It may be an especially useful tool for large-scale multiclass classification problems and lead to more widespread use of SVMs in the engineering community due to its good performance.

Nonparallel Support Vector Machine Based on One Optimization Problem for Pattern Recognition

In this paper,we present a novel nonparallel support vector machine based on one optimization problem(NSVMOOP)for binary classification.Our NSVMOOP is formulated aiming to separate classes from the largest possible angle between the normal vectors and the decision hyperplanes in the feature space,at the same time implementing the structural risk minimization principle.Different from other nonparallel classifiers,such as the representative twin support vector machine,it constructs two nonparallel hyperplanes simultaneously by solving a single quadratic programming problem,on which a modified sequential minimization optimization algorithm is explored.The NSVMOOP is analyzed theoretically and implemented experimentally.Experimental results on both artificial and publicly available benchmark datasets show its feasibility and effectiveness.

Training SVMs on a bound vectors set based on Fisher projection

Standard support vector machines （SVMs） train- ing algorithms have O（l3） computational and O（l2） space complexities, where l is the training set size. It is thus com- /putationally infeasible on very large data sets.To alleviate the Computational burden in SVM training, we propose an algo- rithm to train SVMs on a bound vectors set that is extracted based on Fisher projection. For linear separate problems, we use linear Fisher discriminant to compute the projection line, while for non-linear separate problems, we use kernel Fisher discriminant to compute the projection line. For each case, we select a certain ratio samples whose projections are adja- cent to those of the other class as bound vectors. Theoretical analysis shows that the proposed algorithm is with low com- putational and space complexities.Extensive experiments on several classification benchmarks demonstrate the effective- ness of our approach.

A parallel and scalable digital architecture for training support vector machines

To facilitate the application of support vector machines (SVMs) in embedded systems,we propose and test a parallel and scalable digital architecture based on the sequential minimal optimization (SMO) algorithm for training SVMs.By taking advantage of the mature and popular SMO algorithm,the numerical instability issues that may exist in traditional numerical algorithms are avoided.The error cache updating task,which dominates the computation time of the algorithm,is mapped into multiple processing units working in parallel.Experiment results show that using the proposed architecture,SVM training problems can be solved effectively with inexpensive fixed-point arithmetic and good scalability can be achieved.This architecture overcomes the drawbacks of the previously proposed SVM hardware that lacks the necessary flexibility for embedded applications,and thus is more suitable for embedded use,where scalability is an important concern.

A Stack-based Ensemble Framework for Detecting Cancer MicroRNA Biomarkers

MicroRNA （miRNA） plays vital roles in biological processes like RNA splicing and regulation of gene expression. Studies have revealed that there might be possible links between onco- genesis and expression profiles of some miRNAs, due to their differential expression between normal and tumor tissues. However, the automatic classification of miRNAs into different cate- gories by considering the similarity of their expression values has rarely been addressed. This article proposes a solution framework for solving some real-life classification problems related to cancer, miRNA, and mRNA expression datasets. In the first stage, a mulfiobjective optimization based framework, non-dominated sorting genetic algorithm II, is proposed to automatically determine the appropriate classifier type, along with its suitable parameter and feature combinations, pertinent for classifying a given dataset. In the second page, a stack-based ensemble technique is employed to get a single combinatorial solution from the set of solutions obtained in tke first stage. The performance of the proposed two-stage approach is evaluated on several cancer and RNA expression pro- file datasets. Compared to several state-of-the-art approaches for classifying different datasets, our method shows supremacy in the accuracy of classification.