文献类型：专利

英文题名：Convex boundary based learning sample extraction method, involves building learning sample subset convex center, forming sample index by using quadrant hash table, and extracting boundary sample from quadrant boundaries of convex set

作者：YUAN Y;GU Y;TAN X;RUAN T

机构：[1]UNIV EAST CHINA SCI & TECHNOLOGY

申请号：CN108052592-A

申请日：2017-12-12

公开日：2018-05-18

语种：英文

收录：DERWENT

摘要：NOVELTY - The method involves pre-cutting a data set and pre-processing an input database, where the database comprises a missing value and an abnormal value. A learning sample subset convex center is built to form a convex set. A sample index is formed by using a quadrant hash table. A boundary sample is extracted from quadrant boundaries of the convex set. The sample is deleted by performing normalization operation. A central point is identified. Coordinate transformation is obtained. Content samples are marked in a coordinate of the quadrant. USE - Convex boundary based learning sample extraction method. ADVANTAGE - The method enables reducing number of machine learning training sample and improving machine learning performance by adopting common data classification algorithm to obtain edge samples. DESCRIPTION OF DRAWING(S) - The drawing shows a flowchart illustrating a convex boundary based learning sample extraction method. '(Drawing includes non-English language text)'