Learning Bounded Subsets of Lₚ
نویسندگان
چکیده
We study learning problems in which the underlying class is a bounded subset of L p and target Y belongs to . Previously, minimax sample complexity estimates were known under such boundedness assumptions only when p=∞. present sharp estimate that holds for any p > 4; it based on procedure suited heavy-tailed problems.
منابع مشابه
Hardy approximation to Lp functions on subsets of the circle
We consider approximation of L p functions by Hardy functions on subsets of the circle. We rst derive some properties of traces of Hardy classes on such subsets, and then turn to a generalization of classical extremal problems involving norm constraints on the complementary subset. Approximation dans l'espace de Hardy de fonctions L p sur un sous{ensemble du cercle R esum e : Nous etudions l'ap...
متن کاملCompactness of the Bounded Closed Subsets of E 2 T
For simplicity, we use the following convention: a, b are real numbers, r is a real number, i, j, n are natural numbers, M is a non empty metric space, p, q, s are points of E2 T, e is a point of E2, w is a point of En, z is a point of M , A, B are subsets of En T, P is a subset of E2 T, and D is a non empty subset of E2 T. One can prove the following propositions: (2) a − 2 · a = −a. (3) −a + ...
متن کاملFunctions of bounded variation on compact subsets of the plane
A major obstacle in extending the theory of well-bounded operators to cover operators whose spectrum is not necessarily real has been the lack of a suitable variation norm applicable to functions defined on an arbitrary nonempty compact subset σ of the plane. In this paper we define a new Banach algebra BV (σ) of functions of bounded variation on such a set and show that the function theoretic ...
متن کاملCompactness of the Bounded Closed Subsets of E 2
For simplicity, we adopt the following convention: a, b denote real numbers, r denotes a real number, i, j, n denote natural numbers, M denotes a non empty metric space, p, q, s denote points of E2 T, e denotes a point of E2, w denotes a point of En, z denotes a point of M, A, B denote subsets of En T, P denotes a subset of E2 T, and D denotes a non empty subset of E2 T. One can prove the follo...
متن کاملlp-Norm Multiple Kernel Learning
Learning linear combinations of multiple kernels is an appealing strategy when the right choice of features is unknown. Previous approaches to multiple kernel learning (MKL) promote sparse kernel combinations to support interpretability and scalability. Unfortunately, this l1-norm MKL is rarely observed to outperform trivial baselines in practical applications. To allow for robust kernel mixtur...
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: IEEE Transactions on Information Theory
سال: 2021
ISSN: ['0018-9448', '1557-9654']
DOI: https://doi.org/10.1109/tit.2021.3083553