| 查看: 659 | 回复: 3 | |||
| 当前只显示满足指定条件的回帖，点击这里查看本话题的所有回帖 | |||
[求助]
求润色一段英文摘要，谢谢
|
|||
|
摘要：基于交替非负最小二乘算法的框架，本文提出一种非负矩阵分解的非单调自适应BB（Barzilai-Borwein）步长算法. 虽然该算法的步长不是由线搜索取得的，但是满足非单调线搜索，从而保证了算法的全局收敛性. 同时该算法使用自适应BB步长和梯度的Lipschitz常数来提高算法的收敛速度. 最后在理论上证明了该算法是收敛的，同时数值试验和人脸识别的试验结果表明该算法是有效的且优于其他算法. Abstract: A new algorithm named nonmonotone adaptive Barzilai-Borwein stepsize (MABB) algorithm was proposed for solving the nonnegative matrix factorization. It is based on the alternating nonnegative least squares (ANLS) framework and the stepsize which is not achieved by line search but satisfies the nonmonotone line search, thus ensuring the global convergence of the algorithm. Furthermore, adaptive BB stepsize and the gradient of the Lipschitz constant are used to accelerate convergence. Finally, the algorithm is theoretically proved convergence. At the same time, the test results of numerical experiments and face recognition show that the proposed algorithm has advantages over the existing algorithms in terms of efficiency. |
回复此楼» 猜你喜欢
售SCI一区文章，我:8 O5 51O 54,科目齐全
已经有4人回复
-
售SCI一区文章，我:8 O5 51O 54,科目齐全
已经有4人回复
-
售SCI一区文章，我:8 O5 51O 54,科目齐全
已经有5人回复
-
售SCI一区文章，我:8 O5 51O 54,科目齐全
已经有4人回复
-
售SCI一区文章，我:8 O5 51O 54,科目齐全
已经有6人回复
-
售SCI一区文章，我:8 O5 51O 54,科目齐全
已经有6人回复
-
售SCI一区文章，我:8 O5 51O 54,科目齐全
已经有6人回复
-
售SCI一区文章，我:8 O5 51O 54,科目齐全
已经有7人回复
-
售SCI一区文章，我:8 O5 51O 54,科目齐全
已经有7人回复
-
售SCI一区文章，我:8 O5 51O 54,科目齐全
已经有7人回复
武汉一心一译 
捐助贵宾 (著名写手)
- 翻译EPI: 502
- 应助: 8 (幼儿园)
- 金币: 2283.1
- 散金: 5914
- 红花: 32
- 帖子: 1665
- 在线: 321.9小时
- 虫号: 3587652
- 注册: 2014-12-10
- 性别: GG
- 专业: 高分子合成化学
【答案】应助回帖
商家已经主动声明此回帖可能含有宣传内容|
摘要：基于 交替 非负最小二乘算法 的框架，本文 提出 一种 非负矩阵分解的 非单调自适应BB（Barzilai-Borwein） 步长算法. 虽然 该算法的 步长 不是 由 线搜索 取得的，但是 满足 非单调线搜索，从而 保证了 算法的 全局收敛性. 同时 该算法 使用 自适应 BB步长 和 梯度的Lipschitz常数 来 提高 算法的 收敛速度. 最后在 理论上 证明了 该算法 是收敛的，同时 数值试验 和 人脸识别 的 试验结果 表明 该算法 是有 效的 且 优于 其他算法. Abstract: Based on alternating nonnegative least squares (ANLS) framework, in this paper, we proposed the nonmonotone adaptive BB（Barzilai-Borwein）step-length algorithm to solve nonnegative matrix factorization. Although the step-length of this algorithm was not obtained by line search, it still meet the characteristics of nonmonotone line search, so that the global convergence of the algorithm can be guaranteed. In addition, this algorithm increases the convergence rate by adopting self-adaptive BB step-length and gradient Lipschitz constant. At last, the convergence characteristics of the algorithm was theoretically proved, moreover the experiment results of related numerical experiment and face identification reveals the efficacy of the algorithm as well as its superiority over other algorithms. |
4楼2015-08-13 21:47:59
ssssllllnnnn 
至尊木虫 (知名作家)
Translator and Proofreader
- 翻译EPI: 1690
- 应助: 452 (硕士)
- 金币: 31580.9
- 红花: 100
- 帖子: 7681
- 在线: 19966.6小时
- 虫号: 3328089
- 注册: 2014-07-17
- 专业: 肿瘤发生
【答案】应助回帖
|
写的不错，只要将时态统一了就可以了： Abstract: A new algorithm named nonmonotone adaptive Barzilai-Borwein stepsize (MABB) algorithm was proposed for solving the nonnegative matrix factorization. It WAS based on the alternating nonnegative least squares (ANLS) framework and the stepsize which WAS not achieved by line search but satisfieD the nonmonotone line search, thus ensuring the global convergence of the algorithm. Furthermore, adaptive BB stepsize and the gradient of the Lipschitz constant WERE used to accelerate convergence. Finally, the algorithm WAS theoretically proved convergenT. At the same time, the test results of numerical experiments and face recognition showED that the proposed algorithm haD advantages over the existing algorithms in terms of efficiency. |
2楼2015-08-13 03:38:31
【答案】应助回帖
| Based on the alternating non-negative least squares (ANLS) framework, the paper has proposed a new algorithm named non-monotone adaptive Barzilai-Borwein step-size (MABB) algorithm. The step-size of the algorithm is not calculated through line search but it satisfies the non-monotone line search, ensuring the global convergence of the algorithm. Furthermore, the adaptive BB step-size and the gradient of the Lipschitz constant are also used in the algorithm to accelerate convergence. Finally, the algorithm is theoretically proved convergent and the test results of numerical experiments and face recognition show that the proposed algorithm is effective and outruns other existing algorithms. |
3楼2015-08-13 14:45:18
10