Quantum computer systems do not have that type of separation. Whereas they might embody some quantum reminiscence, the information is usually housed immediately within the qubits, whereas computation includes performing operations, referred to as gates, immediately on the qubits themselves. The truth is, there was an indication that, for supervised machine studying, the place a system can study to categorise gadgets after coaching on pre-classified knowledge, a quantum system can outperform classical ones, even when the information being processed is housed on classical {hardware}.
This type of machine studying depends on what are referred to as variational quantum circuits. It is a two-qubit gate operation that takes a further issue that may be held on the classical facet of the {hardware} and imparted to the qubits by way of the management alerts that set off the gate operation. You’ll be able to consider this as analogous to the communications concerned in a neural community, with the two-qubit gate operation equal to the passing of knowledge between two synthetic neurons and the issue analogous to the burden given to the sign.
That is precisely the system {that a} workforce from the Honda Analysis Institute labored on in collaboration with a quantum software program firm referred to as Blue Qubit.
Pixels to qubits
The main focus of the brand new work was totally on the best way to get knowledge from the classical world into the quantum system for characterization. However the researchers ended up testing the outcomes on two totally different quantum processors.
The issue they had been testing is one in every of picture classification. The uncooked materials was from the Honda Scenes dataset, which has photographs taken from roughly 80 hours of driving in Northern California; the pictures are tagged with details about what’s within the scene. And the query the researchers wished the machine studying to deal with was a easy one: Is it snowing within the scene?
