I am trying to understand how Arbitrary Encoding, as described in this qiskit site https://learn.qiskit.org/course/machine-learning/data-encoding, works and can encode 12 features using only 3 qubits.
I can understand that encoding 2 features per qubit is possible, using Ry and Rz. Neverthless, the example circuit shown below contains 3 qubits and encodes 4 features per qubit using CNOTs in between the Ry and Rz gates.
I would appreciate if someone could shed some light on the reasons for such approach to work.
