Let R be a prime ring and m, n be fixed non-negative integers such that m+n ≠ 0. Suppose L is an (m+m+1)-power closed Lie ideal, and this means ure+n+1 ∈ L for all u ∈ L. If charR = 0 or a prime p 〉 2(m ...Let R be a prime ring and m, n be fixed non-negative integers such that m+n ≠ 0. Suppose L is an (m+m+1)-power closed Lie ideal, and this means ure+n+1 ∈ L for all u ∈ L. If charR = 0 or a prime p 〉 2(m + n), we characterize the additive maps d: L → R satisfying d(um+n+1) = (m -+n + 1)umd(u)un (resp., d(um+n+l) = umd(u)un) for all u ∈ L.展开更多
场景图为描述图像内容的结构图(Graph),其在生成过程中存在两个问题:1)二步式场景图生成方法造成有益信息流失,使得任务难度提高;2)视觉关系长尾分布使得模型发生过拟合、关系推理错误率上升。针对这两个问题,文中提出结合多尺度特征图...场景图为描述图像内容的结构图(Graph),其在生成过程中存在两个问题:1)二步式场景图生成方法造成有益信息流失,使得任务难度提高;2)视觉关系长尾分布使得模型发生过拟合、关系推理错误率上升。针对这两个问题,文中提出结合多尺度特征图和环型关系推理的场景图生成模型SGiF(Scene Graph in Features)。首先,计算多尺度特征图上的每一特征点存在视觉关系的可能性,并将存在可能性高的特征点特征提取出来;然后,从被提取出的特征中解码得到主宾组合,根据解码结果的类别差异,对结果进行去重,以此得到场景图结构;最后,根据场景图结构检测包含目标关系边在内的环路,将环路上的其他边作为计算调整因子的输入,以该因子调整原关系推理结果,并最终完成场景图的生成。实验设置SGGen和PredCls作为验证项,在大型场景图生成数据集VG(Visual Genome)子集上的实验结果表明,通过使用多尺度特征图,相比二步式基线,SGiF的视觉关系检测命中率提升了7.1%,且通过使用环型关系推理,相比非环型关系推理基线,SGiF的关系推理命中率提升了2.18%,从而证明了SGiF的有效性。展开更多
In this paper,we show that every injective Jordan semi-triple multiplicative map on the Hermitian matrices must be surjective,and hence is a Jordan ring isomorphism.
Let Z/(p^e) be the integer residue ring modulo p^e with p an odd prime and integer e ≥ 3. For a sequence a over Z/(p^e), there is a unique p-adic decomposition a- = a-0 +a-1 .p +… + a-e-l .p^e-1 where each a-...Let Z/(p^e) be the integer residue ring modulo p^e with p an odd prime and integer e ≥ 3. For a sequence a over Z/(p^e), there is a unique p-adic decomposition a- = a-0 +a-1 .p +… + a-e-l .p^e-1 where each a-i can be regarded as a sequence over Z/(p), 0 ≤ i ≤ e - 1. Let f(x) be a primitive polynomial over Z/(p^e) and G'(f(x),p^e) the set of all primitive sequences generated by f(x) over Z/(p^e). For μ(x) ∈ Z/(p)[x] with deg(μ(x)) ≥ 2 and gad(1 + deg(μ(x)),p- 1) = 1, setφe-1 (x0, x1,… , xe-1) = xe-1. [μ(xe-2) + ηe-3(x0, X1,…, xe-3)] + ηe-2(x0, X1,…, xe-2) which is a function of e variables over Z/(p). Then the compressing mapφe-1 : G'(f(x),p^e) → (Z/(p))^∞ ,a-→φe-1(a-0,a-1, … ,a-e-1) is injective. That is, for a-,b-∈ G'(f(x),p^e), a- = b- if and only if φe-1 (a-0,a-1, … ,a-e-1) = φe-1(b-0, b-1,… ,b-e-1). As for the case of e = 2, similar result is also given. Furthermore, if functions φe-1 and ψe-1 over Z/(p) are both of the above form and satisfy φe-1(a-0,a-1,…,a-e-1)=ψe-1(b-0, b-1,… ,b-e-1) for a-,b-∈G'(f(x),p^e), the relations between a- and b-, φe-1 and ψe-1 are discussed展开更多
文摘Let R be a prime ring and m, n be fixed non-negative integers such that m+n ≠ 0. Suppose L is an (m+m+1)-power closed Lie ideal, and this means ure+n+1 ∈ L for all u ∈ L. If charR = 0 or a prime p 〉 2(m + n), we characterize the additive maps d: L → R satisfying d(um+n+1) = (m -+n + 1)umd(u)un (resp., d(um+n+l) = umd(u)un) for all u ∈ L.
文摘场景图为描述图像内容的结构图(Graph),其在生成过程中存在两个问题:1)二步式场景图生成方法造成有益信息流失,使得任务难度提高;2)视觉关系长尾分布使得模型发生过拟合、关系推理错误率上升。针对这两个问题,文中提出结合多尺度特征图和环型关系推理的场景图生成模型SGiF(Scene Graph in Features)。首先,计算多尺度特征图上的每一特征点存在视觉关系的可能性,并将存在可能性高的特征点特征提取出来;然后,从被提取出的特征中解码得到主宾组合,根据解码结果的类别差异,对结果进行去重,以此得到场景图结构;最后,根据场景图结构检测包含目标关系边在内的环路,将环路上的其他边作为计算调整因子的输入,以该因子调整原关系推理结果,并最终完成场景图的生成。实验设置SGGen和PredCls作为验证项,在大型场景图生成数据集VG(Visual Genome)子集上的实验结果表明,通过使用多尺度特征图,相比二步式基线,SGiF的视觉关系检测命中率提升了7.1%,且通过使用环型关系推理,相比非环型关系推理基线,SGiF的关系推理命中率提升了2.18%,从而证明了SGiF的有效性。
基金Supported by the National Natural Science Foundation of China (Grant Nos.11001194 10771157)the Natural Science Foundation of Shanxi Province (Grant No.2009021002)
文摘In this paper,we show that every injective Jordan semi-triple multiplicative map on the Hermitian matrices must be surjective,and hence is a Jordan ring isomorphism.
基金Supported by the National Natural Science Foundation of China(60673081)863 Program(2006AA01Z417)
文摘Let Z/(p^e) be the integer residue ring modulo p^e with p an odd prime and integer e ≥ 3. For a sequence a over Z/(p^e), there is a unique p-adic decomposition a- = a-0 +a-1 .p +… + a-e-l .p^e-1 where each a-i can be regarded as a sequence over Z/(p), 0 ≤ i ≤ e - 1. Let f(x) be a primitive polynomial over Z/(p^e) and G'(f(x),p^e) the set of all primitive sequences generated by f(x) over Z/(p^e). For μ(x) ∈ Z/(p)[x] with deg(μ(x)) ≥ 2 and gad(1 + deg(μ(x)),p- 1) = 1, setφe-1 (x0, x1,… , xe-1) = xe-1. [μ(xe-2) + ηe-3(x0, X1,…, xe-3)] + ηe-2(x0, X1,…, xe-2) which is a function of e variables over Z/(p). Then the compressing mapφe-1 : G'(f(x),p^e) → (Z/(p))^∞ ,a-→φe-1(a-0,a-1, … ,a-e-1) is injective. That is, for a-,b-∈ G'(f(x),p^e), a- = b- if and only if φe-1 (a-0,a-1, … ,a-e-1) = φe-1(b-0, b-1,… ,b-e-1). As for the case of e = 2, similar result is also given. Furthermore, if functions φe-1 and ψe-1 over Z/(p) are both of the above form and satisfy φe-1(a-0,a-1,…,a-e-1)=ψe-1(b-0, b-1,… ,b-e-1) for a-,b-∈G'(f(x),p^e), the relations between a- and b-, φe-1 and ψe-1 are discussed