Multi-kernel Approach Research Articles

Hierarchical approach for multiscale support vector regression.

Support vector regression (SVR) is based on a linear combination of displaced replicas of the same function, called a kernel. When the function to be approximated is nonstationary, the single kernel approach may be ineffective, as it is not able to follow the variations in the frequency content in the different regions of the input space. The hierarchical support vector regression (HSVR) model presented here aims to provide a good solution also in these cases. HSVR consists of a set of hierarchical layers, each containing a standard SVR with Gaussian kernel at a given scale. Decreasing the scale layer by layer, details are incorporated inside the regression function. HSVR has been widely applied to noisy synthetic and real datasets and it has shown the ability in denoising the original data, obtaining an effective multiscale reconstruction of better quality than that obtained by standard SVR. Results also compare favorably with multikernel approaches. Furthermore, tuning the SVR configuration parameters is strongly simplified in the HSVR model.

Multikernel Adaptive Filtering

This paper exemplifies that the use of multiple kernels leads to efficient adaptive filtering for nonlinear systems. Two types of multikernel adaptive filtering algorithms are proposed. One is a simple generalization of the kernel normalized least mean square (KNLMS) algorithm [2], adopting a coherence criterion for dictionary designing. The other is derived by applying the adaptive proximal forward-backward splitting method to a certain squared distance function plus a weighted block l1 norm penalty, encouraging the sparsity of an adaptive filter at the block level for efficiency. The proposed multikernel approach enjoys a higher degree of freedom than those approaches which design a kernel as a convex combination of multiple kernels. Numerical examples show that the proposed approach achieves significant gains particularly for nonstationary data as well as insensitivity to the choice of some design-parameters.

Gene Ontology-driven inference of protein–protein interactions using inducers

Protein-protein interactions (PPIs) are pivotal for many biological processes and similarity in Gene Ontology (GO) annotation has been found to be one of the strongest indicators for PPI. Most GO-driven algorithms for PPI inference combine machine learning and semantic similarity techniques. We introduce the concept of inducers as a method to integrate both approaches more effectively, leading to superior prediction accuracies. An inducer (ULCA) in combination with a Random Forest classifier compares favorably to several sequence-based methods, semantic similarity measures and multi-kernel approaches. On a newly created set of high-quality interaction data, the proposed method achieves high cross-species prediction accuracies (Area under the ROC curve ≤ 0.88), rendering it a valuable companion to sequence-based methods. Software and datasets are available at http://bioinformatics.org.au/go2ppi/ m.ragan@uq.edu.au.

Development of an operational procedure to estimate surface albedo from the SEVIRI/MSG observing system by using POLDER BRDF measurements: II. Comparison of several inversion techniques and uncertainty in albedo estimates

This paper develops a concept related to the significant information extracted from the bidirectional reflectance distribution function (BRDF) of the terrestrial targets. The main issues are: the choice of the BRDF model, the solution to the inverse problem, and the accuracy assessment of estimated albedo. The present concept is based on the fact that the exact solution to the inverse problem belongs to a statistically significant region centered on the least squares solution (LSS). Nonetheless, LSS may be useless if the matrix inversion yields an ill-posed problem. It is then recommended to seek an alternative solution, which will yield a similar confidence interval but will be more physically sound. A list of 15 kernels entering in a basic model is examined by means of factor analysis performed in vector space, which is spanned at all known kernels. The application is carried out with synthetic angular data generated for the SEVIRI/MSG observing system. Models are evaluated based on statistical results—minimum of squared sum of residuals (SSR) and maximum of explained variance—after adjustment on reflectance data corresponding to a wide set of land cover types. Since the matrix of the model is almost singular, we identified an optimal subset model consisting of eight kernels, which has higher conditioned index and falls within the 95% confidence interval. It was found that the reflectance predicted by multi-kernel model is consistent with measurements. The idea in opting for a multi-kernel approach comes from the necessity to perform a higher angular resolution for the BRDF retrieval. Inversion experiments confirmed an advantage of the composite model over conventional three-parameter models in accuracy assessment of reflectance and albedo in the case of uniform and restricted angular samplings. Three methods are considered: statistical inversion (provided by the LSS), ridge regression and statistical regularization. The two latter are advised to solve the ill-conditioned inverse problem. Statistical regularization uses a priori statistical information. The inversion numerical experiment with SEVIRI/MSG angular geometry shows that only ridge regression provides a reasonable solution when a composite model is used. In addition, ridge regression and statistical regularization methods provide physically acceptable solutions in terms of BRDF and albedo predictability, even for three-parameter models. It is advised that LSS be implemented only at the middle of the summer season in the Northern Hemisphere. Otherwise, the use of ridge regression and statistical regularization is recommended to retrieve BRDF and albedo at other time periods in extra-tropical latitudes.

Multiagent AI implementations an emerging software engineering trend

A model is presented for the design and implementation of software systems using multiagent artificial intelligence (AI) and concurrent software engineering techniques. The stages of conceptualization, design and implementation are defined by AI methods. It is proposed that software systems be designed through knowledge acquisition, specification, and multiagent implementations. Multiagent implementations are proposed here to facilitate a fault-tolerant software-design methodology, applying recent research that has led to fault-tolerant AI systems. A particular approach to, and an AI formulation for designing fault-free and fault-tolerant software is presented, based on the agent models of computation. An approach using object-oriented design, coupled with a novel multi-kernel approach, is presented. A system is defined by many communicating pairs of kernels, each defining a part of the system, as specified by object-level knowledge acquisition. An overview of agent morphisms and algebras is presented and applied to the design.