In the Marvel Cinematic Universe (MCU), J.A.R.V.I.S. was Tony Stark’s original AI assistant. He helped Mr. Stark at his home, in all of his vehicles, and, of course, in his Iron Man suits. It’s safe to say that without J.A.R.V.I.S. the Iron Man systems would be remarkably less functional than they were ultimately portrayed.

Quantum algorithms and hybrid classical-quantum algorithms usually try to play the role of an Iron Man suit. They tackle problems head on, and try to solve them. Qunova’s HIVQE is a novel algorithm that plays the role of J.A.R.V.I.S. instead.

# Variational Quantum Eigensolver (VQE)

Though often referred to as an algorithm, VQE has become more of a family of algorithms. Its many variations are often found tackling quantum chemistry problems, as well as the occasional optimization problems.

The main problem with VQE is that it doesn’t scale. The quantum circuit, which is in and of itself an artform, includes parameters that are optimized by classical algorithms. Scaling up the problem increases the size of the circuit, which exponentially increases the number of parameters that need to be optimized, which creates an intractable classical machine learning problem. It also exponentially increases the number of measurements that need to be taken, which also makes the problem intractable. That’s two intractable problems for the price of one!

According to Qunova CEO Dr. Kevin Rhee, execution of VQE by classical means cannot scale beyond 30 qubits and VQE with current quantum computers cannot scale beyond 12 qubits. Besides these limited problem sizes, current quantum hardware cannot achieve chemical accuracy and will often find non-optimal solutions. HIVQE is proposed as a much more viable alternative.

# Handover Iterative VQE (HIVQE)

Qunova’s approach greatly simplifies VQE. Instead of allowing the problem space to grow exponentially, near-zero values are discarded. And instead of taking all possible measurements, only one type of measurement is taken. You may be therefore wondering, logically, how the heck HIVQE solves problems.

It doesn’t.

Rather than trying to solve problems directly, the goal is to use a quantum computer to convert classically hard problems into classically tractable problems. In other words, HIVQE assumes the role of J.A.R.V.I.S. and aids a classical algorithm, aka Iron Man, in solving problems. The result of using a quantum computer, therefore, is greater scalability while the result of using a classical algorithm is chemical accuracy.

What’s the future of HIVQE?

The successor to VQE, called quantum phase estimation (QPE), requires quantum circuits that are far too large to run on current quantum computers. What’s really interesting about HIVQE, though, is that it requires far fewer operations ((N^2)/2 two-qubit gates) than QPE. It’ll be interesting to determine the problem size, once QPE can be run, that it’ll become advantageous.

# Conclusion

HIVQE runs faster than VQE because it requires fewer measurements over fewer iterations to optimize its fewer parameters. Qunova reports success with 12-24 qubits, which already surpasses VQE, and claims HIVQE is the first approach to achieve “chemical accuracy.” By feeding into a classical solver, the curve from HIVQE matches the curve you would plot if using a classical solver. Furthermore, a big departure from VQE is waiting 10 minutes for a chemically accurate solution versus waiting hours for an incorrect solution.

Qunova claims HIVQE is 10X more scalable than VQE, and has selected Configuration Interaction (CI) for comparison. The goal is to match the CI limit, which is Si2H6, which has 34 electrons and 152 orbitals. Unfortunately, HIVQE needs a number of qubits that is double the 152 orbitals, and a quantum computer this size doesn’t exist. This is like an after-credits scene in an MCU movie, teasing us with what will come next, but forcing us to wait until the next one. If this analogy is any good, it’s worth noting that J.A.R.V.I.S. never let Iron Man down.

https://shield-files.fandom.com/wiki/J.A.R.V.I.S.