The optimal design of training sequences for channel estimation in multiple-input multiple-output （MIMO） systems under spatially correlated fading is considered. The channel is assumed to be a block-fading model wit...The optimal design of training sequences for channel estimation in multiple-input multiple-output （MIMO） systems under spatially correlated fading is considered. The channel is assumed to be a block-fading model with spatial correlation known at both the transmitter and the receiver. To minimize the channel estimation error, optimal training sequences are designed to exploit full information of the spatial correlation under the criterion of minimum mean square error （MMSE）. It is investigated that the spatial correlation is helpful to decrease the estimation error and the proposed training sequences have good performance via simulations.展开更多
The purpose of this article is to study the weak and strong convergence of implicit iteration process with errors to a common fixed point for a finite family of asymptotically nonexpansive mappings and nonexpansive ma...The purpose of this article is to study the weak and strong convergence of implicit iteration process with errors to a common fixed point for a finite family of asymptotically nonexpansive mappings and nonexpansive mappings in Banach spaces. The results presented in this article extend and improve the corresponding results of [1, 2, 4-9, 11-15].展开更多
针对句子分类任务常面临着训练数据不足,而且文本语言具有离散性,在语义保留的条件下进行数据增强具有一定困难,语义一致性和多样性难以平衡的问题,本文提出一种惩罚生成式预训练语言模型的数据增强方法(punishing generative pre-train...针对句子分类任务常面临着训练数据不足,而且文本语言具有离散性,在语义保留的条件下进行数据增强具有一定困难,语义一致性和多样性难以平衡的问题,本文提出一种惩罚生成式预训练语言模型的数据增强方法(punishing generative pre-trained transformer for data augmentation,PunishGPT-DA)。设计了惩罚项和超参数α,与负对数似然损失函数共同作用微调GPT-2(generative pre-training 2.0),鼓励模型关注那些预测概率较小但仍然合理的输出;使用基于双向编码器表征模型(bidirectional encoder representation from transformers,BERT)的过滤器过滤语义偏差较大的生成样本。本文方法实现了对训练集16倍扩充,与GPT-2相比,在意图识别、问题分类以及情感分析3个任务上的准确率分别提升了1.1%、4.9%和8.7%。实验结果表明,本文提出的方法能够同时有效地控制一致性和多样性需求,提升下游任务模型的训练性能。展开更多
随着大规模预训练语言模型的出现,文本生成技术已取得突破性进展。然而,在开放性文本生成领域,生成的内容缺乏拟人化的情感特征,使生成的文本难以让人产生共鸣和情感上的联系,可控文本生成在弥补当前文本生成技术不足方面具有重要意义...随着大规模预训练语言模型的出现,文本生成技术已取得突破性进展。然而,在开放性文本生成领域,生成的内容缺乏拟人化的情感特征,使生成的文本难以让人产生共鸣和情感上的联系,可控文本生成在弥补当前文本生成技术不足方面具有重要意义。首先,在ChnSentiCorp数据集的基础上完成主题和情感属性的扩展,同时,为构建一个可生成流畅文本且情感丰富的多元可控文本生成模型,提出一种基于扩散序列的可控文本生成模型DiffuSeq-PT。该模型以扩散模型为基础架构,利用主题情感属性和文本数据在无分类器引导条件下对序列执行扩散过程,使用预训练模型ERNIE 3.0(Large-scale Knowledge Enhanced Pre-training for Language Understanding and Generation)的编码解码能力贴合扩散模型的加噪去噪过程,最终生成符合相关主题和多情感粒度的目标文本。与基准模型DiffuSeq相比,所提模型在2个公开的真实数据集(ChnSentiCorp和辩论数据集)上分别取得0.13和0.01的BERTScore值的提升,困惑度分别下降了14.318和9.46。展开更多
基金the National Science Foundation for Distinguished Young Scholars (60725105)the SixthProject of the Key Project of National Nature Science Foundation of China (60496316)+2 种基金the National "863" Project (2007AA012288)the National Nature Science Foundation of China (60572146)the "111" Project (B08038).
文摘The optimal design of training sequences for channel estimation in multiple-input multiple-output （MIMO） systems under spatially correlated fading is considered. The channel is assumed to be a block-fading model with spatial correlation known at both the transmitter and the receiver. To minimize the channel estimation error, optimal training sequences are designed to exploit full information of the spatial correlation under the criterion of minimum mean square error （MMSE）. It is investigated that the spatial correlation is helpful to decrease the estimation error and the proposed training sequences have good performance via simulations.
基金The present studies were supported by the Natural Science Foundation of Zhe-jiang Province (Y605191)the Natural Science Foundation of Heilongjiang Province (A0211)the Key Teacher Creating Capacity Fund of Heilongjiang General College (1053G015)the Scientific Research Foundation from Zhejiang Province Education Committee (20051897)the Starting Foundation of Scientific Research from Hangzhou Teacher's College.
文摘The purpose of this article is to study the weak and strong convergence of implicit iteration process with errors to a common fixed point for a finite family of asymptotically nonexpansive mappings and nonexpansive mappings in Banach spaces. The results presented in this article extend and improve the corresponding results of [1, 2, 4-9, 11-15].
文摘针对句子分类任务常面临着训练数据不足,而且文本语言具有离散性,在语义保留的条件下进行数据增强具有一定困难,语义一致性和多样性难以平衡的问题,本文提出一种惩罚生成式预训练语言模型的数据增强方法(punishing generative pre-trained transformer for data augmentation,PunishGPT-DA)。设计了惩罚项和超参数α,与负对数似然损失函数共同作用微调GPT-2(generative pre-training 2.0),鼓励模型关注那些预测概率较小但仍然合理的输出;使用基于双向编码器表征模型(bidirectional encoder representation from transformers,BERT)的过滤器过滤语义偏差较大的生成样本。本文方法实现了对训练集16倍扩充,与GPT-2相比,在意图识别、问题分类以及情感分析3个任务上的准确率分别提升了1.1%、4.9%和8.7%。实验结果表明,本文提出的方法能够同时有效地控制一致性和多样性需求,提升下游任务模型的训练性能。
文摘随着大规模预训练语言模型的出现,文本生成技术已取得突破性进展。然而,在开放性文本生成领域,生成的内容缺乏拟人化的情感特征,使生成的文本难以让人产生共鸣和情感上的联系,可控文本生成在弥补当前文本生成技术不足方面具有重要意义。首先,在ChnSentiCorp数据集的基础上完成主题和情感属性的扩展,同时,为构建一个可生成流畅文本且情感丰富的多元可控文本生成模型,提出一种基于扩散序列的可控文本生成模型DiffuSeq-PT。该模型以扩散模型为基础架构,利用主题情感属性和文本数据在无分类器引导条件下对序列执行扩散过程,使用预训练模型ERNIE 3.0(Large-scale Knowledge Enhanced Pre-training for Language Understanding and Generation)的编码解码能力贴合扩散模型的加噪去噪过程,最终生成符合相关主题和多情感粒度的目标文本。与基准模型DiffuSeq相比,所提模型在2个公开的真实数据集(ChnSentiCorp和辩论数据集)上分别取得0.13和0.01的BERTScore值的提升,困惑度分别下降了14.318和9.46。
