ABSTRACT
Let Π https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4114.tif"/> be an NP-hard optimization problem, and let A https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4115.tif"/> be an approximation algorithm for Π https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4116.tif"/> . Given an instance I https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4117.tif"/> of Π https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4118.tif"/> , denote by A ( I ) https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4119.tif"/> the objective value obtained by A https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4120.tif"/> for I https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4121.tif"/> , and by O P T ( I ) https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4122.tif"/> the optimal objective value. The approximation ratio of A https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4123.tif"/> for the instance I https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4124.tif"/> is R A ( I ) = A ( I ) / O P T ( I ) https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4125.tif"/> when Π https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4126.tif"/> is a minimization problem, and R A ( I ) = O P T ( I ) / A ( I ) https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4127.tif"/> when Π https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4128.tif"/> is a maximization problem; thus, R A ( I ) ≥ 1 https://s3-euw1-ap-pe-df-pch-content-public-u.s3.eu-west-1.amazonaws.com/9781351236423/7ef10423-eda9-42b3-a9ea-761fc7500b5b/content/equ_4129.tif"/>
