Jinbo Bi

MILES: Multiple-Instance Learning via Embedded Instance Selection

Multiple-instance problems arise from the situations where training class labels are attached to sets of samples (named bags), instead of individual samples within each bag (called instances). Most previous multiple-instance learning (MIL) algorithms are developed based on the assumption that a bag is positive if and only if at least one of its instances is positive. Although the assumption works well in a drug activity prediction problem, it is rather restrictive for other applications, especially those in the computer vision area. We propose a learning method, MILES (multiple-instance learning via embedded instance selection), which converts the multiple-instance learning problem to a standard supervised learning problem that does not impose the assumption relating instance labels to bag labels. MILES maps each bag into a feature space defined by the instances in the training bags via an instance similarity measure. This feature mapping often provides a large number of redundant or irrelevant features. Hence, 1-norm SVM is applied to select important features as well as construct classifiers simultaneously. We have performed extensive experiments. In comparison with other methods, MILES demonstrates competitive classification accuracy, high computation efficiency, and robustness to labeling uncertainty

Active learning via transductive experimental design

This paper considers the problem of selecting the most informative experiments x to get measurements y for learning a regression model y = f(x). We propose a novel and simple concept for active learning, transductive experimental design, that explores available unmeasured experiments (i.e., unlabeled data) and has a better scalability in comparison with classic experimental design methods. Our in-depth analysis shows that the new method tends to favor experiments that are on the one side hard-to-predict and on the other side representative for the rest of the experiments. Efficient optimization of the new design problem is achieved through alternating optimization and sequential greedy search. Extensive experimental results on synthetic problems and three real-world tasks, including questionnaire design for preference learning, active learning for text categorization, and spatial sensor placement, highlight the advantages of the proposed approaches.

Bayesian multiple instance learning: automatic feature selection and inductive transfer

We propose a novel Bayesian multiple instance learning (MIL) algorithm. This algorithm automatically identifies the relevant feature subset, and utilizes inductive transfer when learning multiple (conceptually related) classifiers. Experimental results indicate that the proposed MIL method is more accurate than previous MIL algorithms and selects a much smaller set of useful features. Inductive transfer further improves the accuracy of the classifier as compared to learning each task individually.

Column-generation boosting methods for mixture of kernels

We devise a boosting approach to classification and regression based on column generation using a mixture of kernels. Traditional kernel methods construct models based on a single positive semi-definite kernel with the type of kernel predefined and kernel parameters chosen according to cross-validation performance. Our approach creates models that are mixtures of a library of kernel models, and our algorithm automatically determines kernels to be used in the final model. The 1-norm and 2-norm regularization methods are employed to restrict the ensemble of kernel models. The proposed method produces sparser solutions, and thus significantly reduces the testing time. By extending the column generation (CG) optimization which existed for linear programs with 1-norm regularization to quadratic programs with 2-norm regularization, we are able to solve many learning formulations by leveraging various algorithms for constructing single kernel models. By giving different priorities to columns to be generated, we are able to scale CG boosting to large datasets. Experimental results on benchmark data are included to demonstrate its effectiveness.

A Geometric Approach to Support Vector Regression

We develop an intuitive geometric framework for support vector regression (SVR). By examining when †-tubes exist, we show that SVR can be regarded as a classification problem in the dual space. Hard and soft †-tubes are constructed by separating the convex or reduced convex hulls respectively of the training data with the response variable shifted up and down by †. A novel SVR model is proposed based on choosing the max-margin plane between the two shifted datasets. Maximizing the margin corresponds to shrinking the eective †-tube. In the proposed approach, the eects of the choices of all parameters become clear geometrically. The kernelized model corresponds to separating the convex or reduced convex hulls in feature space. Generalization bounds for classification can be extended to characterize the generalization performance of the proposed approach. We propose a simple iterative nearest-point algorithm that can be directly applied to the reduced convex hull case in order to construct soft †-tubes. Computational comparisons with other SVR formulations are also included.

A sparse support vector machine approach to region-based image categorization

Automatic image categorization using low-level features is a challenging research topic in computer vision. In this paper, we formulate the image categorization problem as a multiple-instance learning (MIL) problem by viewing an image as a bag of instances, each corresponding to a region obtained from image segmentation. We propose a new solution to the resulting MIL problem. Unlike many existing MIL approaches that rely on the diverse density framework, our approach performs an effective feature mapping through a chosen metric distance function. Thus the MIL problem becomes solvable by a regular classification algorithm. Sparse SVM is adopted to dramatically reduce the regions that are needed to classify images. The selected regions by a sparse SVM approximate to the target concepts in the traditional diverse density framework. The proposed approach is a lot more efficient in computation and less sensitive to the class label uncertainty. Experimental results are included to demonstrate the effectiveness and robustness of the proposed method.

An Improved Multi-task Learning Approach with Applications in Medical Diagnosis

We propose a family of multi-task learning algorithms for collaborative computer aided diagnosis which aims to diagnose multiple clinically-related abnormal structures from medical images. Our formulations eliminate features irrelevant to all tasks, and identify discriminative features for each of the tasks. A probabilistic model is derived to justify the proposed learning formulations. By equivalence proof, some existing regularization-based methods can also be interpreted by our probabilistic model as imposing a Wishart hyperprior. Convergence analysis highlights the conditions under which the formulations achieve convexity and global convergence. Two real-world medical problems: lung cancer prognosis and heart wall motion analysis, are used to validate the proposed algorithms.

Efficient model selection for regularized linear discriminant analysis

Classical Linear Discriminant Analysis (LDA) is not applicable for small sample size problems due to the singularity of the scatter matrices involved. Regularized LDA (RLDA) provides a simple strategy to overcome the singularity problem by applying a regularization term, which is commonly estimated via cross-validation from a set of candidates. However, cross-validation may be computationally prohibitive when the candidate set is large. An efficient algorithm for RLDA is presented that computes the optimal transformation of RLDA for a large set of parameter candidates, with approximately the same cost as running RLDA a small number of times. Thus it facilitates efficient model selection for RLDA.An intrinsic relationship between RLDA and Uncorrelated LDA (ULDA), which was recently proposed for dimension reduction and classification is presented. More specifically, RLDA is shown to approach ULDA when the regularization value tends to zero. That is, RLDA without any regularization is equivalent to ULDA. It can be further shown that ULDA maps all data points from the same class to a common point, under a mild condition which has been shown to hold for many high-dimensional datasets. This leads to the overfitting problem in ULDA, which has been observed in several applications. Thetheoretical analysis presented provides further justification for the use of regularization in RLDA. Extensive experiments confirm the claimed theoretical estimate of efficiency. Experiments also show that, for a properly chosen regularization parameter, RLDA performs favorably in classification, in comparison with ULDA, as well as other existing LDA-based algorithms and Support Vector Machines (SVM).

A fast iterative algorithm for fisher discriminant using heterogeneous kernels

We propose a fast iterative classification algorithm for Kernel Fisher Discriminant (KFD) using heterogeneous kernel models. In contrast with the standard KFD that requires the user to predefine a kernel function, we incorporate the task of choosing an appropriate kernel into the optimization problem to be solved. The choice of kernel is defined as a linear combination of kernels belonging to a potentially large family of different positive semidefinite kernels. The complexity of our algorithm does not increase significantly with respect to the number of kernels on the kernel family. Experiments on several benchmark datasets demonstrate that generalization performance of the proposed algorithm is not significantly different from that achieved by the standard KFD in which the kernel parameters have been tuned using cross validation. We also present results on a real-life colon cancer dataset that demonstrate the efficiency of the proposed method.

Joint Optimization of Cascaded Classifiers for Computer Aided Detection

The existing methods for offline training of cascade classifiers take a greedy search to optimize individual classifiers in the cascade, leading inefficient overall performance. We propose a new design of the cascaded classifier where all classifiers are optimized for the final objective function. The key contribution of this paper is the AND-OR framework for learning the classifiers in the cascade. In earlier work each classifier is trained independently using the examples labeled as positive by the previous classifiers in the cascade, and optimized to have the best performance for that specific local stage. The proposed approach takes into account the fact that an example is classified as positive by the cascade if it is labeled as positive by all the stages and it is classified as negative if it is rejected at any stage in the cascade. An offline training scheme is introduced based on the joint optimization of the classifiers in the cascade to minimize an overall objective function. We apply the proposed approach to the problem of automatically detecting polyps from multi-slice CT images. Our approach significantly speeds up the execution of the computer aided detection (CAD) system while yielding comparable performance with the current state-of-the-art, and also demonstrates favorable results over cascade AdaBoost both in terms of performance and online execution speed.