Covering Numbers for Semicontinuous Functions
Royset, Johannes O.
MetadataShow full item record
Considering the metric space of extended real-valued lower semicontinuous functions under the epi-distance, the paper gives an upper bound on the covering numbers of bounded subsets of such functions. No assumptions about continuity, smoothness, variation, and even niteness of the functions are needed. The bound is shown to be nearly sharp through the construction of a set of functions with covering numbers deviating from the upper bound only by a logarithmic factor. The analogy between lower and upper semicontinuous functions implies that identical covering numbers hold for bounded sets of the latter class of functions as well, but now under the hypo-distance metric.
This paper is in review.
Showing items related by title, author, creator and subject.
Studies in Ship Survivability. I - Ducking the Punch or Absorbing the Blow? Letting the Anticipated Threat Drive Future Warship Design. Studies in Ship Survivability. I - Fighting Dirty? Sustaining Operations in an NBC Environment. Harney, Robert C. (Monterey, California. Naval Postgraduate SchoolInstitute for Joint Warfare Analysis, 1999-11-07);The threat projected by recent novel "Red Team" studies is analyzed to determine its impact on future warship design. The future threat is expected to employ massive numbers of cruise missiles, mines, and submarines ...
In-line and transverse forces on smooth and sand-roughened cylinders in oscillatory flow at high Reynolds numbers Sarpkaya, Turgut (Monterey, California. Naval Postgraduate School, 1976-06-04); NPS-69SL76062This report presents the results of an extensive experimental investigation of the in-line and transverse forces acting on smooth and sand- roughened circular cylinders placed in oscillatory flow at Reynolds numbers up to ...
Knudstrup, Timothy A. (Monterey, California. Naval Postgraduate School, 2007-12);Numeric Function Generators (NFGs) have allowed computation of difficult mathematical functions in less time and with less hardware than commonly employed methods. They compute piecewise linear (or quadratic) approximations ...