However, our kernels, designed to weigh information according to what is considered a priori to be relevant (molecular structure features in the fingerprints, differentially expressed genes in profiles, known protein-protein interactions), have shown poor results when used to train kernel classifiers.
To compute the train-train and train-test kernel matrices, we used the parameter free Χ 2 kernel for all features.
Qiu and Noble [ 20] integrated PPI, protein sequences, gene expression, interologs, and functional information, to train kernel-based models, and achieved high classification accuracy in predicting co-complex protein pairs.
When training any kernel method (e.g. an SVM or a kernel regression method) with this kernel, the resulting function is a weighted sum of kernel evaluations (1).
In the offline training phase, kernel density, which is optional, is incorporated in two stages: vocabulary tree construction and image histogram representation; in the online recognition phase for the query image, the kernel density is only involved in the histogram calculation.
The cost and gamma are parameters used in model training and kernel function of LIBSVM and need to be to tuned for optimal prediction results.
Nowadays, SVMs with string kernels can be trained efficiently on millions of DNA sequences even for large orders K (e.g. Sonnenburg et al., 2007a, thereby inducing enormous feature spaces (for instance, K=30 gives rise to more than 4>10 k-mers).
In this study, different SVMs and RVMs were trained varying kernel and error functions with different set of independent variables.
Values on the x axis correspond to SVM classifiers trained with kernel K (see text) as projected into the subspace spanned by its first x Principal Components.
Representative methods for multi-view learning can be categorized into three groups: co-training, multiple kernel learning, and subspace learning [52].
