摘要:
本文通过连续可微的非凸函数所形成的概率约束,来分析概率约束问题。描述了潜在的概率函数的水平集的切锥和法锥。进一步,基于p-有效点的概念,形成这些问题的一阶和二阶最优性条件。对于离散分布函数的这种情况,产生一个基于修正的指数函数的对偶算法来解决概率约束问题。In this paper, the problem of probability constraints is analyzed by means of the probability constraints formed by continuously differentiable non-convex functions. The tangent and normal cones of the level set of potential probability functions are described. Further, based on the concept of p-efficient points, the first and second order optimality conditions of these problems are formed. For this case of the discrete distribution function, a dual algorithm based on the modified exponential function is generated to solve the probability constraint problem.
摘要:
针对多功能视频编码(Versatile Video Coding,VVC)标准中跨通道线性预测模型(Cross-Component Linear Model,CCLM)无法很好地拟合色度与亮度之间的非线性对应关系这一不足,提出了一种基于注意力机制卷积神经网络的VVC色度预测算法。该算法主要思想是在进行色度预测时,使用对应亮度块的信息与待预测色度块上方与左方的信息作为参考信息输入进卷积神经网络,利用注意力机制对参考信息中的亮度与色度间的内在联系进行分配权重后输入预测网络。实验结果表明,相较于VVC标准算法U分量和V分量的平均码率节省分别为0.64%和0.68%,有效提升了VVC编码性能。