We consider several complexity measures which capture the difficulty of learning under the i.i.d. assumption. Among these measures are growth function and VC dimension, covering number and fat-shattering dimension, and Rademacher complexity from statistical learning theory. Relationships among these complexity measures, their connection to learning, and tools for bounding them are provided. For each complexity measure, a uniform upper bound on the generalization error of classification problems is presented.
CITATION STYLE
V’yugin, V. V. (2015). VC dimension, fat-shattering dimension, rademacher averages, and their applications. In Measures of Complexity: Festschrift for Alexey Chervonenkis (pp. 57–74). Springer International Publishing. https://doi.org/10.1007/978-3-319-21852-6_6
Mendeley helps you to discover research relevant for your work.