登入選單

返回Google圖書搜尋

On Learning to Recognize 3D Objects from Examples

Mekhon Ṿaitsman le-madaʻ. Department of Applied Mathematics and Computer ScienceShimon Edelman

出版	Weizmann Institute of Science, Department of Applied Mathematics and Computer Science, 1991
URL	http://books.google.com.hk/books?id=CTSLHAAACAAJ&hl=&source=gbs_api

註釋

Abstract: "Previous analyses of visual concept acquisition in the context of computational learning theory tended to rely on the unrealistic assumption that concepts are represented and learned as sets of pixels [1]. In comparison, two recently proposed algorithms that learn to recognize three-dimensional objects from examples [2, 3] employ receptive fields for coarse features as the basic unit of representation. This report uses an approach developed by Haussler [4] to show that, under a feature-based definition of complexity, recognition is PAC-learnable from a feasible number of examples in a distribution-free manner."