I encountered this paper on travel time estimation using Tensor Decomposition and at Page 4, Figure 5 there is an algorithm to decompose a tensor using Gradient Descent.
The first line of this algorithm initializes six low-rank matrices. I don't understand how is it possible to initialize these matrices if their dimensions is not known in advance (they are not in the Input section).
In the aforementioned paper, it is not mentioned how the dimensions are computed and I am stumbling around in the dark at the moment.
I am trying to understand something from this tutorial, but it is quite hard to understand for me and if you could at least point me to something more readable would be awesome.
Thanks to everyone willing to help.