منابع مشابه
Interpolation with Variably Scaled Kernels
Within kernel–based interpolation and its many applications, it is a well–documented but unsolved problem to handle the scaling or the shape parameter. We consider native spaces whose kernels allow us to change the kernel scale of a d–variate interpolation problem locally, depending on the requirements of the application. The trick is to define a scale function c on the domain Ω ⊂ Rd to transfo...
متن کاملInterpolating functions with gradient discontinuities via Variably Scaled Kernels
In kernel–based methods, how to handle the scaling or the choice of the shape parameter is a well– documented but still an open problem. The shape or scale parameter can be tuned by the user according to the applications, and it plays a crucial role both for the accuracy of the method and its stability. In [7], the Variably Scaled Kernels (VSKs) were introduced. The idea is to vary the scale in...
متن کاملConvex Deep Learning via Normalized Kernels
Deep learning has been a long standing pursuit in machine learning, which until recently was hampered by unreliable training methods before the discovery of improved heuristics for embedded layer training. A complementary research strategy is to develop alternative modeling architectures that admit efficient training methods while expanding the range of representable structures toward deep mode...
متن کاملLearning Kernels via Margin-and-Radius Ratios
Despite the great success of SVM, it is usually difficult for users to select suitable kernels for SVM classifiers. Kernel learning has been developed to jointly learn both a kernel and an SVM classifier [1]. Most existing kernel learning approaches, e.g., [2, 3, 4], employ the margin based formulation, equivalent to: mink,w,b,ξi 1 2‖w‖ 2 + C ∑ i ξi, s.t. yi〈φ(xi; k), w〉+ b+ ξi ≥ 1, ξi ≥ 0, (1)...
متن کاملLearning Gaussian Process Kernels via Hierarchical Bayes
We present a novel method for learning with Gaussian process regression in a hierarchical Bayesian framework. In a first step, kernel matrices on a fixed set of input points are learned from data using a simple and efficient EM algorithm. This step is nonparametric, in that it does not require a parametric form of covariance function. In a second step, kernel functions are fitted to approximate...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Advances in Computational Mathematics
سال: 2021
ISSN: 1019-7168,1572-9044
DOI: 10.1007/s10444-021-09875-6