欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
为了提高演化硬件（EHW）分类系统的泛化能力和减少硬件代价，提出了一种用于DNA微阵列数据分类的演化硬件多分类器选择性集成学习方法。重点讨论了基于Bagging的选择性集成学习策略和基于虚拟可重构结构的演化硬件分类系统构架。通过对原始数据训练集的随机重采样生成训练子集完成对演化硬件基分类器的训练，并选择其中识别率较高的基分类器进行集成以获得更高的分类性能。演化硬件分类系统对DNA微阵列数据的学习与分类均在Xilinx Virtexxcv2000E FPGA硬件平台上实现。通过对急性白血病和肺癌数据集的对比实验表明：相对于传统演化硬件集成学习方法，这种方法在保证较高识别率的基础上有效降低了硬件代价，且具有更短的学习时间和较强的泛化能力。
英文摘要：
In order to improve the generalization ability and reduce the hardware cost of evolvable hardware （EHW） classification systems, a bagging-based selective ensemble learning method using EMW multiple classifiers was proposed for the classification of DNA microarray data. A bagging-based selective ensemble learning strategy and a virtual reconfigurable architecture-based EHW classification system were studied. In the system learning process, several training subsets were generated by using random sampling from the original training set. The final EHW classifier was built by using the evolved base classifiers with the high classification rate. Both the system learning and the system classification of the EHW for the classification of microarray data were implemented on a Xilinx Virtex xcv2000E FPGA. The comparison of the experimental results of acute leukemia and lung dataset showed the proposed method＇ s advantages of much lower hardware cost, higher recognition rate, shorter learning time and generalization ability compared with traditional EHW ensemble learning schemes.
同期刊论文项目
同项目期刊论文
期刊信息
位置：