Dimension reduction in the search for online bin packing policies
Dimension reduction in the search for online bin packing policies
复制标题
在线装箱政策搜索中的降维
DOI：
10.1145/2464576.2464620
复制
发表时间：
2013
期刊：
影响因子：
--
通讯作者：
Asta S
中科院分区：
文献类型：
--
作者：
Asta S
在在线装箱问题中，必须找到一种策略，根据物品的大小，在物品到达时立即分配到具有已知初始容量的箱子中。在Ozcan和Parkes（GECCO 2011）之前的工作中，策略被表示为二维“矩阵”（数组），然后使用遗传算法（GA）进化出好的矩阵。在这里，我们考虑一种形式的降维，其中矩阵中的变量被分组为从一维向量中提取的元素。我们发现，与正确的分组形式，遗传算法通常会发现这样的“矢量政策”显着更快，但遭受的整体质量损失不大。
In online bin-packing problems, a policy must be found for assigning items, according their size, immediately upon their arrival to bins with known initial capacities. In previous work of Ozcan and Parkes (GECCO 2011), a policy was represented as a 2-dimensional "matrix" (array) and good matrices were then evolved using a genetic algorithm (GA). Here, we consider a form of dimensional reduction in which variables in the matrix are grouped into elements taken from one-dimensional vectors. We find that with the right form of grouping, the GA then typically finds such "vector policies" significantly more quickly, and yet suffers little loss of overall quality.
用于生成启发式的策略矩阵演化
影响因子：
6.2
作者：
E. Özcan;A. Parkes
通讯作者：
A. Parkes