کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
10347289 | 699116 | 2011 | 10 صفحه PDF | دانلود رایگان |
عنوان انگلیسی مقاله ISI
Faster integer-feasibility in mixed-integer linear programs by branching to force change
دانلود مقاله + سفارش ترجمه
دانلود مقاله ISI انگلیسی
رایگان برای ایرانیان
موضوعات مرتبط
مهندسی و علوم پایه
مهندسی کامپیوتر
علوم کامپیوتر (عمومی)
پیش نمایش صفحه اول مقاله

چکیده انگلیسی
Branching in mixed-integer (or integer) linear programming requires choosing both the branching variable and the branching direction. This paper develops a number of new methods for making those two decisions either independently or together with the goal of reaching the first integer-feasible solution quickly. These new methods are based on estimating the probability of satisfying a constraint at the child node given a variable/direction pair. The surprising result is that the first integer-feasible solution is usually found much more quickly when the variable/direction pair with the smallest probability of satisfying the constraint is chosen. This is because this selection forces change in many candidate variables simultaneously, leading to an integer-feasible solution sooner. Extensive empirical results are given.
ناشر
Database: Elsevier - ScienceDirect (ساینس دایرکت)
Journal: Computers & Operations Research - Volume 38, Issue 8, August 2011, Pages 1143-1152
Journal: Computers & Operations Research - Volume 38, Issue 8, August 2011, Pages 1143-1152
نویسندگان
Jennifer Pryor, John W. Chinneck,