This paper studies a novel paradigm for learning formal languages from positive and negative examples which consists of mapping strings to an appropriate highdimensional feature s...
In supervised kernel methods, it has been observed that the performance of the SVM classifier is poor in cases where the diagonal entries of the Gram matrix are large relative to ...
A successful class of image denoising methods is based on Bayesian approaches working in wavelet representations. The performance of these methods improves when relations among th...
Valero Laparra, Juan Gutierrez, Gustavo Camps-Vall...
Kernel methods have been widely studied in the field of pattern recognition. These methods implicitly map, "the kernel trick," the data into a space which is more approp...
We present a novel approach to recognizing Textual nt. Structural features are constructed from abstract tree descriptions, which are automatically extracted from syntactic depend...