Continuing from last time, we will be talking more about tensor notations.
n-mode Vectors
Let be an tensor. An -mode vector of is an -vector obtained from by fixing indices but let vary.
Example 1: Let be a tensor with , , , , , , , and .
Fixing , , and letting vary, we get the 3-mode vector .
Fixing , , and letting vary, we get the 1-mode vector .
Fixing , , and letting vary, we get the 2-mode vector .
Matrix Unfolding
Matrix unfolding is a way to represent a tensor as a matrix. The concept is easy to grasp, but tedious to describe mathematically. So let us start with some examples on 3-mode tensors.
Let be an tensor. It has three different matrix unfoldings and , corresponding to its three modes. has rows and columns where the columns are the 1-mode vectors of . The columns are ordered so that the index varies faster than . So, the first column corresponds to the 1-mode vector where . The second column corresponds to and . The -st column corresponds to and . In general, the 1-mode vector with and appears as the -th column.
Analogously, has rows and columns. Here, the 2-mode vector with and appears as the -th column. Notice that varies faster than . In general, when we say that (or any other unfolded matrices) is of size , we mean that (the number to the right) varies faster than (the number of the left).
A similar situation goes for . It has row and columns with index varies more slowly than .
Example 2: Let be the tensor in Example 1. The three matrix unfoldings of are given below:
In general, a matrix unfolding of an tensor along the -mode is an matrix whose columns are the -mode vectors, and the columns are arranged such that the index varies the fastest and the slowest. To be pedantic, we can specify exactly where the element appears in . It appears on Row and on Column