Topic
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.
Papers published on a yearly basis
Papers
More filters
01 Oct 2019
TL;DR: The presented hierarchical method with nonuniform sampling can improve the final classification precision more greatly for the small-size samples of the field objects and can give a higher overall accuracy and average accuracy comparing to the single spectral classification algorithms and the non-hierarchical LMKL method.
Abstract: Aiming at the high precision classification of hyperspectral image in the case of small-size samples, a hierarchical localized multiple kernel learning (HLMKL) method for hyperspectral image classification is presented in this paper. By using a non-uniform sampling, the hyperspectral image can be hierarchically represented as several data sub-sets with different scales. On this basis, the dimension-reduced features are obtained through the KPCA method, and the hierarchical features and the multiple kernel function can be fused after the respectively mapping by introducing the localized multiple kernel model (LMKL). Ultimately, an effective classifier for hyperspectral image is trained by global optimization learning. Experiment results show that the proposed classification method has better universal practicability and robustness, which can give a higher overall accuracy (OA) and can greatly improve the average accuracy (AA) comparing to the single spectral classification algorithms and the non-hierarchical LMKL method. Furthermore, the presented hierarchical method with nonuniform sampling can improve the final classification precision more greatly for the small-size samples of the field objects.
15 Dec 2009
TL;DR: This paper proposes two methods to classify sports news photos into one of the given six sports categories and to discriminate in-play photos from not-in-play ones using the two-step and one-step methods.
Abstract: In this paper, we treat with in-play classification of sports news photos as an instance of researches on more sophisticated search methods for large-scale photo news databases. We propose two methods to classify sports news photos into one of the given six sports categories and to discriminate in-play photos from not-in-play ones. One is the two-step method which classifies sports categories first and recognizes in-play conditions next, and the other is the one-step method which classifies them simultaneously. In the proposed methods, we integrate textual features extracted from news articles and image features extracted from photo images by Multiple Kernel Learning (MKL). In the experiment of the two-step method, we obtained 99.33% as the classification rate for the sports category classification which is the first step and 80.75% for the in-play classification which is the second step. On the other hand, in the experiment of the one-step method, we obtained 77.08% which was a little less than the result by the two-step method.
17 Jun 2022
TL;DR: In this article , the Kreın-SVM was proposed and developed for AMP classification and prediction by employing the Levenshtein distance and local alignment score as sequence similarity functions.
Abstract: Antimicrobial peptides (AMPs) represent a potential solution to the growing problem of antimicrobial resistance, yet their identification through wet-lab experiments is a costly and time-consuming process. Accurate computational predictions would allow rapid in silico screening of candidate AMPs, thereby accelerating the discovery process. Kernel methods are a class of machine learning algorithms that utilise a kernel function to transform input data into a new representation. When appropriately normalised, the kernel function can be regarded as a notion of similarity between instances. However, many expressive notions of similarity are not valid kernel functions, meaning they cannot be used with standard kernel methods such as the support-vector machine (SVM). The Kreın-SVM represents a generalisation of the standard SVM that admits a much larger class of similarity functions. In this study, we propose and develop Kreın-SVM models for AMP classification and prediction by employing the Levenshtein distance and local alignment score as sequence similarity functions. Utilising two datasets from the literature, each containing more than 3000 peptides, we train models to predict general antimicrobial activity. Our best models achieve an AUC of 0.967 and 0.863 on the test sets of each respective dataset, outperforming the in-house and literature baselines in both cases. We also curate a dataset of experimentally validated peptides, measured against Staphylococcus aureus and Pseudomonas aeruginosa, in order to evaluate the applicability of our methodology in predicting microbe-specific activity. In this case, our best models achieve an AUC of 0.933 and 0.917, respectively. Models to predict both general and microbe-specific activities are made available as web applications.
01 Jan 2013
TL;DR: There is a need for a novel and a more sophisticated approach to a data-driven and problem-oriented kernel design for addressing the problem of the function reconstruction inside/outside the scope of given data points.
Abstract: The paper is concerned with an adaptive kernel design for addressing the problem of the function reconstruction inside/outside the scope of given data points. We analyze the state of the art methods and, by doing that, we show a need for a novel and a more sophisticated approach to a data-driven and problem-oriented kernel design. Finally, we present such approach and show its superiority with respect to the known methods on the numerical experiments with real data.