Conference proceeding
Online estimation for finding a near-maximum value in a large list of numerical data
The Institute of Electrical and Electronics Engineers, Inc. (IEEE) Conference Proceedings
01 Jan 2018
Abstract
Conference Title: 2018 52nd Annual Conference on Information Sciences and Systems (CISS) Conference Start Date: 2018, March 21 Conference End Date: 2018, March 23 Conference Location: Princeton, NJ, USA It is often of interest to find a maximum or near-maximum value in a large (unsorted) list of numerical data (perhaps large values are the object of study, or they help establish the data's range). For large datasets it is not feasible to sort the data, therefore such values must be found by repeatedly querying the dataset via a search algorithm. Compounding the problem is that identifying an index as having near-maximum value appears to require knowledge of the maximum value itself, which may be unknown. This paper proposes an online estimation approach to this problem, assuming knowledge of the dataset's order and its distribution family. We leverage results in extreme value theory (EVT) to estimate the maximum value, and thereby identify a target sample size. The algorithm's performance is evaluated on several synthetic datasets; we show the algorithm performs well, despite the limited information it requires.
Metrics
8 Record Views
Details
- Title
- Online estimation for finding a near-maximum value in a large list of numerical data
- Creators
- Jonathan StokesSteven Weber
- Publication Details
- The Institute of Electrical and Electronics Engineers, Inc. (IEEE) Conference Proceedings
- Publisher
- The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
- Resource Type
- Conference proceeding
- Language
- English
- Academic Unit
- Electrical and Computer Engineering
- Identifiers
- 991019170154404721