A structure and training algorithm for quasi-diagonal recurrent neural network (QDRNN) is presented.
提出一种准对角递归神经网络(QDRNN)结构及学习算法。
2
We introduce concepts of diagonalquasi-convexity and quasi-concavity in hyperconvex metric spaces.
我们介绍了超凸度量空间中对角拟凸和拟凹的概念。
3
We introduced the concept of block directed edge cover diagonalquasi dominant matrix, obtained a new nonsingularity criteria for matrices and distribution theorem on eigenvalues of matrix.