Rotation matrices vs neural_networks for translation

Hello
I wonder why not use a neural net inplace of a rotation matrix for word translation as proposed int the NLP specialization C1_W4.
I’ll be happy to receive some enlightment about this topic.
Hermes

Hi @Hermes_Morales_Gross

I think, it is to illustrate how the simplest “neural network” (just a single matrix) can “translate” from one language to another - given vector of one language embedding “rotates” to another language embedding vector.

Later in the course, of course, we use deeper and more sophisticated architectures for translation application. But first, students have to understand the basic concepts of linear transformation.

Cheers