We use the approach of “optimal” switching to design the adaptive control because the design among multiple models is intuitively more practically feasible than the traditional adaptive control in improving the perf...We use the approach of “optimal” switching to design the adaptive control because the design among multiple models is intuitively more practically feasible than the traditional adaptive control in improving the performances. We prove that for a typical class of nonlinear systems disturbed by random noise, the multiple model adaptive switching control based on WLS (Weighted Least Squares) or projected-LS (Least Squares) is stable and convergent.展开更多
台区电力工单记录反映了台区运行工况和用户需求,是制定台区用电安全管理制度和满足台区用户用电需求的重要依据。针对台区电力工单高复杂性和强专业性给台区工单分类带来的难题,提出一种融合标签平滑(LS)与预训练语言模型的台区电力工...台区电力工单记录反映了台区运行工况和用户需求,是制定台区用电安全管理制度和满足台区用户用电需求的重要依据。针对台区电力工单高复杂性和强专业性给台区工单分类带来的难题,提出一种融合标签平滑(LS)与预训练语言模型的台区电力工单分类模型(MiniRBT-LSTM-GAT)。首先,利用预训练模型计算电力工单文本中的字符级特征向量表示;其次,采用双向长短期记忆网络(BiLSTM)捕捉电力文本序列中的依赖关系;再次,通过图注意力网络(GAT)聚焦对文本分类贡献大的特征信息;最后,利用LS改进损失函数以提高模型的分类精度。所提模型与当前主流的文本分类算法在农网台区电力工单数据集(RSPWO)、浙江省95598电力工单数据集(ZJPWO)和THUCNews(TsingHua University Chinese News)数据集上的实验结果表明,与电力审计文本多粒度预训练语言模型(EPAT-BERT)相比,所提模型在RSPWO、ZJPWO上的查准率和F1值分别提升了2.76、2.02个百分点和1.77、1.40个百分点;与胶囊神经网络模型BRsyn-caps(capsule network based on BERT and dependency syntax)相比,所提模型在THUCNews数据集上的查准率和准确率分别提升了0.76和0.71个百分点。可见,所提模型有效提升了台区电力工单分类的性能,并在THUCNews数据集上表现良好,验证了模型的通用性。展开更多
In this paper, a new method based on LS-SVM (Least Squares Support Vector Machines) is presented to deal with credit assessment in commercial banks for solving the problem of inadequate samples of the financial data,w...In this paper, a new method based on LS-SVM (Least Squares Support Vector Machines) is presented to deal with credit assessment in commercial banks for solving the problem of inadequate samples of the financial data,which usually happended in most banks in China.On the basis of SLT(Statistical Learning Theory),this approach with methodology of SRM (Structural Risk Minimization)will overcome the shortcomings of traditional credit assessment models,such as over fitting and local optimization,and,by using kernel functions in model,it will effectively solve the problems of linear inseparability and selecting parameters of model.The approach has some good properties including a generalization ability and global optimization in terms of sample processing.It is a new way for the credit assessment on the condition of small samples from bank data.The feasibility,effectiveness and practicability of presented approach was verified by experiments.展开更多
基金This work was supported by the National Natural Science Foundation of China.
文摘We use the approach of “optimal” switching to design the adaptive control because the design among multiple models is intuitively more practically feasible than the traditional adaptive control in improving the performances. We prove that for a typical class of nonlinear systems disturbed by random noise, the multiple model adaptive switching control based on WLS (Weighted Least Squares) or projected-LS (Least Squares) is stable and convergent.
文摘台区电力工单记录反映了台区运行工况和用户需求,是制定台区用电安全管理制度和满足台区用户用电需求的重要依据。针对台区电力工单高复杂性和强专业性给台区工单分类带来的难题,提出一种融合标签平滑(LS)与预训练语言模型的台区电力工单分类模型(MiniRBT-LSTM-GAT)。首先,利用预训练模型计算电力工单文本中的字符级特征向量表示;其次,采用双向长短期记忆网络(BiLSTM)捕捉电力文本序列中的依赖关系;再次,通过图注意力网络(GAT)聚焦对文本分类贡献大的特征信息;最后,利用LS改进损失函数以提高模型的分类精度。所提模型与当前主流的文本分类算法在农网台区电力工单数据集(RSPWO)、浙江省95598电力工单数据集(ZJPWO)和THUCNews(TsingHua University Chinese News)数据集上的实验结果表明,与电力审计文本多粒度预训练语言模型(EPAT-BERT)相比,所提模型在RSPWO、ZJPWO上的查准率和F1值分别提升了2.76、2.02个百分点和1.77、1.40个百分点;与胶囊神经网络模型BRsyn-caps(capsule network based on BERT and dependency syntax)相比,所提模型在THUCNews数据集上的查准率和准确率分别提升了0.76和0.71个百分点。可见,所提模型有效提升了台区电力工单分类的性能,并在THUCNews数据集上表现良好,验证了模型的通用性。
文摘In this paper, a new method based on LS-SVM (Least Squares Support Vector Machines) is presented to deal with credit assessment in commercial banks for solving the problem of inadequate samples of the financial data,which usually happended in most banks in China.On the basis of SLT(Statistical Learning Theory),this approach with methodology of SRM (Structural Risk Minimization)will overcome the shortcomings of traditional credit assessment models,such as over fitting and local optimization,and,by using kernel functions in model,it will effectively solve the problems of linear inseparability and selecting parameters of model.The approach has some good properties including a generalization ability and global optimization in terms of sample processing.It is a new way for the credit assessment on the condition of small samples from bank data.The feasibility,effectiveness and practicability of presented approach was verified by experiments.
