In this study, Saccharomyces cerevisiae (baker's yeast) was produced in a fed-batch bioreactor at the optimal dissolved oxygen concentration (DOC) and growth medium temperature. However, it is very difficult to co...In this study, Saccharomyces cerevisiae (baker's yeast) was produced in a fed-batch bioreactor at the optimal dissolved oxygen concentration (DOC) and growth medium temperature. However, it is very difficult to control the DOC using conventional controllers because of the poorly understood and constantly changing dynamics of the bioprocess. A generalized predictive controller (GPC) based on a nonlinear autoregressive integrated moving average exogenous (NARIMAX) model is presented to stabilize the DOC by manipulation of air flow rate. The NARIMAX model is built by an improved recursive least-squares support vector machine, which is trained by an in-place computation scheme and avoids the computation of the inverse of a large matrix and memory reallocation. The proposed nonlinear GPC algorithm requires little preliminary knowledge of the fermentation process, and directly obtains the nonlinear model in matrix form by using iterative multiple modeling instead of linearization at each sampling period. By application of an on-line bioreactor control, experimental results demonstrate the robustness, effectiveness and advantages of the new controller.展开更多
该文针对疫苗接种的相关微博评论进行情感倾向分析,首先利用基于神经网络的Doc2vec模型训练文本向量,继而使用支持向量机(SVM)、随机森林(RF)、逻辑回归(LR)三种机器学习的算法完成情感分类任务,且分别讨论了三种算法在四种不同的Doc2ve...该文针对疫苗接种的相关微博评论进行情感倾向分析,首先利用基于神经网络的Doc2vec模型训练文本向量,继而使用支持向量机(SVM)、随机森林(RF)、逻辑回归(LR)三种机器学习的算法完成情感分类任务,且分别讨论了三种算法在四种不同的Doc2vec模型设定方案下的分类表现。其中Distributed Memory version of Paragraph Vector (PV-DM)算法训练的文本向量中,RF表现最优,在方案一与方案二上其F1分数值均为最高,分别为87.24%、87.50%。基于Distributed Bag of Words version of Paragraph Vector (PV-DBOW)算法训练的文本向量中,SVM表现最优,在方案三与方案四上其F1分数值达到最高,分别为84.11%、83.91%。展开更多
为了更有效地获得缺陷报告的非结构化信息的特征,提出一种D_BBAS(Doc2vec and BERT BiLSTM-attention similarity)方法,它基于大规模缺陷报告库训练特征提取模型,生成能反映深层次语义信息的缺陷摘要文本表示集和缺陷描述文本表示集;利...为了更有效地获得缺陷报告的非结构化信息的特征,提出一种D_BBAS(Doc2vec and BERT BiLSTM-attention similarity)方法,它基于大规模缺陷报告库训练特征提取模型,生成能反映深层次语义信息的缺陷摘要文本表示集和缺陷描述文本表示集;利用这两个分布式的表示集计算出缺陷报告对的相似度,从而得到两个新的相似度特征;这两个新特征将与基于结构化信息生成的传统特征结合后参与重复缺陷报告的检测。在著名开源项目Eclipse、NetBeans和Open Office的缺陷报告库上验证了D_BBAS方法的有效性,其中包含超过50万个缺陷报告。实验结果表明,相比于代表性方法,该方法的F1值平均提升了1.7%,证明了D_BBAS方法的有效性。展开更多
基金Supported by the National Natural Science Foundation of China (20476007, 20676013)
文摘In this study, Saccharomyces cerevisiae (baker's yeast) was produced in a fed-batch bioreactor at the optimal dissolved oxygen concentration (DOC) and growth medium temperature. However, it is very difficult to control the DOC using conventional controllers because of the poorly understood and constantly changing dynamics of the bioprocess. A generalized predictive controller (GPC) based on a nonlinear autoregressive integrated moving average exogenous (NARIMAX) model is presented to stabilize the DOC by manipulation of air flow rate. The NARIMAX model is built by an improved recursive least-squares support vector machine, which is trained by an in-place computation scheme and avoids the computation of the inverse of a large matrix and memory reallocation. The proposed nonlinear GPC algorithm requires little preliminary knowledge of the fermentation process, and directly obtains the nonlinear model in matrix form by using iterative multiple modeling instead of linearization at each sampling period. By application of an on-line bioreactor control, experimental results demonstrate the robustness, effectiveness and advantages of the new controller.
文摘该文针对疫苗接种的相关微博评论进行情感倾向分析,首先利用基于神经网络的Doc2vec模型训练文本向量,继而使用支持向量机(SVM)、随机森林(RF)、逻辑回归(LR)三种机器学习的算法完成情感分类任务,且分别讨论了三种算法在四种不同的Doc2vec模型设定方案下的分类表现。其中Distributed Memory version of Paragraph Vector (PV-DM)算法训练的文本向量中,RF表现最优,在方案一与方案二上其F1分数值均为最高,分别为87.24%、87.50%。基于Distributed Bag of Words version of Paragraph Vector (PV-DBOW)算法训练的文本向量中,SVM表现最优,在方案三与方案四上其F1分数值达到最高,分别为84.11%、83.91%。
