Multiple kernel learning

About: Multiple kernel learning is a research topic. Over the lifetime, 1630 publications have been published within this topic receiving 56082 citations.

Simple but effective methods for combining kernels in computational biology

Abstract: Complex biological data generated from various experiments are stored in diverse data types in multiple datasets. By appropriately representing each biological dataset as a kernel matrix then combining them in solving problems, the kernel-based approach has become a spotlight in data integration and its application in bioinformatics and other fields as well. While linear combination of unweighed multiple kernels (UMK) is popular, there have been effort on multiple kernel learning (MKL) where optimal weights are learned by semi-definite programming or sequential minimal optimization (SMO-MKL). These methods provide high accuracy of biological prediction problems, but very complicated and hard to use, especially for non-experts in optimization. These methods are also usually of high computational cost and not suitable for large data sets. In this paper, we propose two simple but effective methods for determining weights for conic combination of multiple kernels. The former is to learn optimal weights formulated by our measure FSM for kernel matrix evaluation (feature space-based kernel matrix evaluation measure), denoted by FSM-MKL. The latter assigns a weight to each kernel that is proportional to the quality of the kernel, determining by direct cross validation, named proportionally weighted multiple kernels (PWMK). Experimental comparative evaluation of the four methods UMK, SMO-MKL, FSM-MKL and PWMK for the problem of protein-protein interactions shows that our proposed methods are simpler, more efficient but still effective. They achieved performances almost as high as that of MKL and higher than that of UMK.

Nonlinear Low-Rank Matrix Completion for Human Motion Recovery

Abstract: Human motion capture data has been widely used in many areas, but it involves a complex capture process and the captured data inevitably contains missing data due to the occlusions caused by the actor’s body or clothing. Motion recovery, which aims to recover the underlying complete motion sequence from its degraded observation, still remains as a challenging task due to the nonlinear structure and kinematics property embedded in motion data. Low-rank matrix completion-based methods have shown promising performance in short-time-missing motion recovery problems. However, low-rank matrix completion, which is designed for linear data, lacks the theoretic guarantee when applied to the recovery of nonlinear motion data. To overcome this drawback, we propose a tailored nonlinear matrix completion model for human motion recovery. Within the model, we first learn a combined low-rank kernel via multiple kernel learning. By exploiting the learned kernel, we embed the motion data into a high dimensional Hilbert space where motion data is of desirable low-rank and we then use the low-rank matrix completion to recover motions. In addition, we add two kinematic constraints to the proposed model to preserve the kinematics property of human motion. Extensive experiment results and comparisons with five other state-of-the-art methods demonstrate the advantage of the proposed method.

Localized Multiple Kernel Learning Via Sample-Wise Alternating Optimization

Abstract: Our objective is to train support vector machines (SVM)-based localized multiple kernel learning (LMKL), using the alternating optimization between the standard SVM solvers with the local combination of base kernels and the sample-specific kernel weights. The advantage of alternating optimization developed from the state-of-the-art MKL is the SVM-tied overall complexity and the simultaneous optimization on both the kernel weights and the classifier. Unfortunately, in LMKL, the sample-specific character makes the updating of kernel weights a difficult quadratic nonconvex problem. In this paper, starting from a new primal-dual equivalence, the canonical objective on which state-of-the-art methods are based is first decomposed into an ensemble of objectives corresponding to each sample, namely, sample-wise objectives. Then, the associated sample-wise alternating optimization method is conducted, in which the localized kernel weights can be independently obtained by solving their exclusive sample-wise objectives, either linear programming (for l1-norm) or with closed-form solutions (for lp-norm). At test time, the learnt kernel weights for the training data are deployed based on the nearest-neighbor rule. Hence, to guarantee their generality among the test part, we introduce the neighborhood information and incorporate it into the empirical loss when deriving the sample-wise objectives. Extensive experiments on four benchmark machine learning datasets and two real-world computer vision datasets demonstrate the effectiveness and efficiency of the proposed algorithm.

Modern Machine Learning Techniques and Their Applications in Cartoon Animation Research

Abstract: The integration of machine learning techniques and cartoon animation research is fast becoming a hot topic. This book helps readers learn the latest machine learning techniques, including patch alignment framework; spectral clustering, graph cuts, and convex relaxation; ensemble manifold learning; multiple kernel learning; multiview subspace learning; and multiview distance metric learning. It then presents the applications of these modern machine learning techniques in cartoon animation research. With these techniques, users can efficiently utilize the cartoon materials to generate animations in areas such as virtual reality, video games, animation films, and sport simulations

Localized Multiple Kernel Regression

Abstract: Multiple kernel learning (MKL) uses a weighted combination of kernels where the weight of each kernel is optimized during training. However, MKL assigns the same weight to a kernel over the whole input space. Our main objective is the formulation of the localized multiple kernel learning (LMKL) framework that allows kernels to be combined with different weights in different regions of the input space by using a gating model. In this paper, we apply the LMKL framework to regression estimation and derive a learning algorithm for this extension. Canonical support vector regression may over fit unless the kernel parameters are selected appropriately; we see that even if provide more kernels than necessary, LMKL uses only as many as needed and does not overfit due to its inherent regularization.