Hybrid Quantum Classical Algorithms

2332 1

I was preparing slides (around 9pm on Feb 12th, sipping a super nice California Cabernet blend, from Napa, as I am now 😊, writing this) for my talk in CMU Physics Department the next day when I received this email:

I read with interest your pre-print on the arXiv this evening. 

For those folks more interested in wine than in quantum computing:

Del Dotto Winery.

The Beast.

Totally Awesome.

Ok, back to quantum computing.

We had posted our third quantum computing paper just a few hours ago!

It is titled:

Graver Bases via Quantum Annealing with Application to Non-Linear Integer Programs

As you know, I have been working on quantum computing with Raouf and Hedayat for some time now.

We call ourselves the CMU Quantum Computing Group.

Continuing, this email went:

I am working on an optical annealing machine at Stanford that does not suffer from the connectivity issues of D-Wave’s system (we see a ~10-million-times performance difference between our system and D-Wave’s on dense problems).

What? 10-million times! Right.

This looks a bit too good to be true, I thought.

Thanks to NASA, USRA and Google, I have access to a D-Wave quantum computer to test my algorithms.

Our paper describes a novel hybrid quantum-classical approach and summarizes some of our computational findings (on D-Wave).

Fake news permeates quantum computing stuff these days and so I was about to delete the email.

But then I saw that this was an email from Stanford, so I kept reading:

However, we do also suffer from limited precision in our machine.

Aha! Some reality here.

He continues:

We have not perfectly calibrated how limited it is, but I think it’s likely around 5 bits. (We are working on improved versions of the machine where the precision might reach 8 bits in the future.) I would however be very curious to try a few of your instances on my machine.

My talk (part of a special seminar series on quantum initiatives) was titled:

Quantum Computing and Integer Optimization: An Overview

If you are curious about my talk, here is a video.

Warning! It is about an hour long, and discusses (among other things):

Ising model, spectral gap, computational algebraic geometry, toric ideals, randomized algorithms, graph theory, test sets, kernel of matrices.

I did not have time to really get into Morse homology and Gauss-Bonnet theorem, though, so you know, and are not disappointed.

In any case, you may enjoy the first 10 minutes as I discuss why I am so interested in quantum computing!

Back to the email. I knew he had read our paper quite carefully:

Would you mind sending me some sample instances of Problem 4.4.1 (in Ising form – as J matrices and h vectors)? I think it would be interesting for us to try an instance with just 20 spins, and then one with 50 spins, and finally one with 100 spins. I think running this with various choices of t could be informative; maybe starting with t=1, t=10, and t=100. (If the experiment works, and we get a publishable result, you’d of course be co-authors on any resulting publication, needless to say.)

We had a great zoom session earlier today, and indeed we are going to check out how well our instances are tackled by his machine.

Stay tuned!

In the spirit of transparency, this was was not the only email I received on our work as I was creating my slides.

This one stood out:

Dear professor:

I have gone through your work on quantum adiabatic evolution.

I am working on the same problem and trying to develop a better quantum adiabatic algorithm… May I request a discussion on skype, if you agree.

I would be thankful to you for your kind cooperation for the same.

Who is he?

A senior researcher, from Joint Institute of Nuclear Research, from, wait for it:

Dubna, Moscow Oblast, Russia.

Probably not a good idea to Skype?


1 comment

  1. Hi Sridhar … looks interesting … both the Stanford guy and the Beast. Cheers, Kathy

Leave a Reply