决策树报文分类算法

报文分类是网络的基本功能,研究人员在过去二十年提出了众多解决方案,其中决策树报文分类算法由于吞吐量高、适用于多字段、可流水线化等特点受到了广泛关注和深入研究。本文介绍了决策树算法最新研究成果,阐述了决策树报文分类算法的几何意义、常用技术和测试基准,从节点切割技术和规则集分组技术两个维度对决策树算法进行了系统分析和归纳。针对两类常用的决策树构建技术介绍了其中的典型算法,对比了各种典型算法的设计思路和特点,分析了它们的适用场景。总结并展望了决策树算法的下一步研究方向。     

上下文感知的深度弱监督图像哈希表示学习方法

针对现有深度监督图像哈希表示学习方法依赖于图像的类别信息,难以在现实中被广泛应用问题,利用与图像相关的标签信息作为监督信息,提出上下文感知的深度弱监督图像哈希表示学习方法。该方法一方面通过自适应捕获图像区域特征的相关上下文来增强它们的表示能力,另一方面通过引入判别损失来提高学习到的哈希码表示的判别性。在现有两个公开数据集上的大量实验结果证明了该方法的有效性。     

基于不稳定性采样的主动学习方法

传统的主动学习方法往往仅基于当前的目标模型来挑选样本,而忽略了历史模型所蕴含的对未标注样本预测稳定性的信息。因此,提出基于不稳定性采样的主动学习方法,依据历史模型的预测差异来估计每个未标注样本对提高模型性能的潜在效用。该方法基于历史模型对样本的预测后验概率之间的差异来衡量无标注样本的不稳定性,并挑选最不稳定的样本进行查询。在多个数据集上的大量实验结果验证了方法的有效性。     

异步非平稳干扰抑制方法

为解决单通道条件下异步非平稳干扰抑制问题,提出基于数据驱动的稀疏分量分析干扰抑制方法,旨在从接收到的混叠信号中恢复期望信号。该方法利用深度卷积神经网络对输入/输出端数据间的复杂映射关系的强大建模能力,实现了目标信号稀疏域的自适应选择、稀疏域中目标信号稀疏表示的自适应学习以及目标信号的自动恢复。与以往干扰抑制算法不同,所提方法在时域上完成了“端到端”的信号波形恢复,且对混叠观测无先验要求,相比现有方法更具普适性。仿真实验验证了所提干扰抑制方法在不同环境噪声和干扰信号强度及泛化测试条件下的有效性,对干扰的抑制性能显著优于现有算法。     

Bicriteria scheduling on a single batching machine with job transportation and deterioration considerations

We study a single batching machine scheduling problem with transportation and deterioration considerations arising from steel production. A set of jobs are transported, one at a time, by a vehicle from a holding area to the single batching machine. The machine can process several jobs simultaneously as a batch. The processing time of a job will increase if the duration from the time leaving the holding area to the start of its processing exceeds a given threshold. The time needed to process a batch is the longest of the job processing times in the batch. The problem is to determine the job sequence for transportation and the job batching for processing so as to minimize the makespan and the number of batches. We study four variations (P1, P2, P3, P4) of the problem with different treatments of the two criteria. We prove that all the four variations are strongly NP‐hard and further develop polynomial time algorithms for their special cases. For each of the first three variations, we propose a heuristic algorithm and analyze its worst‐case performance. For P4, which is to find the Pareto frontier, we provide a heuristic algorithm and an exact algorithm based on branch and bound. Computational experiments show that all the heuristic algorithms perform well on randomly generated problem instances, and the exact algorithm for P4 can obtain Pareto optimal schedules for small‐scale instances. © 2014 Wiley Periodicals, Inc. Naval Research Logistics 61: 269–285, 2014     

Proactive release procedures for just‐in‐time job shop environments,subject to machine failures

In this paper, we consider just‐in‐time job shop environments (job shop problems with an objective of minimizing the sum of tardiness and inventory costs), subject to uncertainty due to machine failures. We present techniques for proactive uncertainty management that exploit prior knowledge of uncertainty to build competitive release dates, whose execution improves performance. These techniques determine the release dates of different jobs based on measures of shop load, statistical data of machine failures, and repairs with a tradeoff between inventory and tardiness costs. Empirical results show that our methodology is very promising in comparison with simulated annealing and the best of 39 combinations of dispatch rules & release policies, under different frequencies of breakdowns. We observe that the performance of the proactive technique compared to the other two approaches improves in schedule quality (maximizing delivery performance while minimizing costs) with increase in frequency of breakdowns. The proactive technique presented here is also computationally less expensive than the other two approaches. © 2004 Wiley Periodicals, Inc. Naval Research Logistics, 2004     

An FPTAS for scheduling a two‐machine flowshop with one unavailability interval

We study a deterministic two‐machine flowshop scheduling problem with an assumption that one of the two machines is not available in a specified time period. This period can be due to a breakdown, preventive maintenance, or processing unfinished jobs from a previous planning horizon. The problem is known to be NP‐hard. Pseudopolynomial dynamic programming algorithms and heuristics with worst case error bounds are given in the literature to solve the problem. They are different for the cases when the unavailability interval is for the first or second machine. The existence of a fully polynomial time approximation scheme (FPTAS) was formulated as an open conjecture in the literature. In this paper, we show that the two cases of the problem under study are equivalent to similar partition type problems. Then we derive a generic FPTAS for the latter problems with O(n⁵/ε⁴) time complexity. © 2003 Wiley Periodicals, Inc. Naval Research Logistics, 2004.     

试论英语学习中词典的使用

词典是汇集词语进行解释、为人们释疑解惑的工具书。初级英语学习者主要依靠教师授课进行学习,而中高级学习者有更多的时间利用英语词典进行自学。词典是英语学习者的必备工具书,但学习者未能充分利用。因此,有必要对以英语学习词典为主的工具书的使用进行探讨,从而得出结论:要使学习者从词典中充分获益,有必要对英语学习者进行词典使用方面的训练。     

探讨研究生学习动机与培养质量的相关性

以往研究研究生的学习动机侧重于量化研究,本文试图从质性研究的角度对当代研究生学习动机进行探讨,归纳出不同的学习类型,分析当代研究生学习动机与培养质量的关系,以及目前由于学习动机造成的培养质量问题,以提高研究生培养质量。     

Visual-attention gabor filter based online multi-armored target tracking

The multi-armored target tracking (MATT) plays a crucial role in coordinated tracking and strike. The occlusion and insertion among targets and target scale variation is the key problems in MATT. Most state-of-the-art multi-object tracking (MOT) works adopt the tracking-by-detection strategy, which rely on compute-intensive sliding window or anchoring scheme in detection module and neglect the target scale variation in tracking module. In this work, we proposed a more efficient and effective spatial-temporal attention scheme to track multi-armored target in the ground battlefield. By simulating the structure of the retina, a novel visual-attention Gabor filter branch is proposed to enhance detection. By introducing temporal information, some online learned target-specific Convolutional Neural Networks (CNNs) are adopted to address occlusion. More importantly, we built a MOT dataset for armored targets, called Armored Target Tracking dataset (ATTD), based on which several comparable experiments with state-of-the-art methods are conducted. Experimental results show that the proposed method achieves outstanding tracking performance and meets the actual application requirements.