Quantum Bayesian Networks

July 19, 2014

Yikes, How Will This Movie End?

Filed under: Uncategorized — rrtucci @ 5:43 pm

"Boys, it wasn't the airplanes that got him [D-Wave], twas beauty [the gate model quantum computer] that killed the beast." (last line of the prescient 1933 movie, "King Kong")

Boys, it wasn’t the airplanes that got him [D-Wave], twas beauty [gate model quantum computers] that killed the beast.– last line of the prescient 1933 movie, “King Kong”.

I am fairly confident that DWave will implement a more versatile Hamiltonian soon, but that won’t solve the problem they have that they are operating at a temperature above the energy gap. –  Peter Shor, Jun 24, 2014, source of quote

… they [D-Wave] use qubits with very short coherence times of the order of a few nanoseconds, while the total time to perform one annealing run is 20 microseconds. The qubits are thus coherent for only a fraction of the total time, and this raises the question whether they are “coherent enough” or “quantum enough” to show a quantum speedup. –Matthias Troyer, June 19 2014, source of quote

July 16, 2014

Gate Model and D-Wave Quantum Computers Go Mano a Mano Over Machine Learning

Filed under: Uncategorized — rrtucci @ 4:24 am

Check out the following exciting new paper

Bayesian Network Structure Learning Using Quantum Annealing,
by Bryan A. O’Gorman, Alejandro Perdomo-Ortiz, Ryan Babbush, Alan Aspuru-Guzik and Vadim Smelyanskiy

I haven’t read it yet but I like very much the idea behind it.

The authors have been honest and cited my paper proposing doing the same thing (learning the structure of a classical Bayesian Network) but using a different kind of quantum computer. Whereas my paper is for doing this using a gate model QC, theirs is for doing it using a quantum annealer (viz. D-Wave).

Wouldn’t it be great if a horse race were to ensue between the quantum gaters and quantum annealers to see who can do B net learning better!!! A gauntlet has been thrown down. Yikes, I say. Once more unto the breach, dear friends, once more.

July 15, 2014

Six Californias, Says D-Wave

Filed under: Uncategorized — rrtucci @ 7:25 pm

quantum computer
Check out the following news item

‘Six Californias’ plan may make 2016 ballot
(by Laura Mandaro, USA TODAY Network , July 15, 2014)


SAN FRANCISCO — A plan backed by venture capitalist Tim Draper to split California into six states has gained enough signatures to make the November 2016 ballot, the plan’s backers say.
A Twitter account belonging to the nonprofit Six Californias tweeted on Monday that “#SixCalifornias will be submitting signatures in Sacramento tomorrow for placement on the November 2016 ballot. Stay tuned for coverage!”

Draper is a founding member of the venture capital firm Draper Fisher Jurvetson, known for its investments in successful growth ventures such as Hotmail, Baidu, Tesla Motors and Skype. Recently, Draper won the federal government’s auction of bitcoins once owned by online drugs portal Silk Road.

The relevance of this news item to quantum computing might not be immediately obvious to the casual observer so let me be the first to explain it.

Tim Draper is a billionaire who has made his fortune mainly from web-based companies. His VC firm Draper-Fisher-Jurvetson is one of the main investors in D-Wave. He got the idea of splitting California into six parts during a D-Wave session, which I’ve copy & pasted below

> Hello Tim
(%i1) How do I cure cancer?
(%o1) 6
(%i2) What is wrong with this Lockheed F-35 (Joint Strike Fighter) software program?
(%o2) 6
(%i3) How do I stop climate change?
(%o3) 6
(%i4) How do I fix all of California's problems?
(%o4) 6
(%i5) How many billionaires should rule California?
(%o5) 6

July 13, 2014

The Romantic Side of D-Wave

Filed under: Uncategorized — rrtucci @ 5:07 am

“Oh, dearest Leonardo, it’s so romantic of you to court me in sight of the D-Wave”
“Don’t worry sweet Kate. This ship will never sink because it is being steered by a D-Wave. 160 million dollars and the finest Burnaby Canada minds have gone into its construction.”

July 12, 2014

Betting On the Future of D-Wave

Filed under: Uncategorized — rrtucci @ 7:56 pm

Check out the following news item:

D-Wave Systems raises $30M to keep commercializing its quantum computer
By Derrick Harris (gigaom.com, Jul. 10, 2014)

My own reaction to this story is, Goldman-Sachs, Harris-Draper-Fisher-Jurvetson, and Canadian Banks, you guys should join gamblers anonymous. Face it, you have a serious gambling addiction.

In light of the now famous paper by Troyer et al which concludes there is no discernible SCALING advantage over classical computers for the 512 qubit D-wave chip, any cool headed, rational investor would have waited until the results come in for D-Wave’s 1000 qubit machine, promised to arrive by the end of this year, before adding another 30 million dollars to their already considerable bet. But gambling addicts are not rational. When they are already losing big, instead of walking away, they just bet their home and their children’s college fund, because they are SURE their luck is going to change in the next round.

In my opinion, dumb venture capitalist companies like Harris-Draper-Fisher-Jurvetson and greedy, unscrupulous investment bankers like Goldman-Sachs and their new partners in crime, Canadian Banks, should be investing only a fraction of their ill-gotten money in D-wave and the remaining fraction in gate model quantum computers. But instead, they are putting all their eggs, 160 million dollars worth, in one, D-Wave basket. How prudent is that for an investment strategy? And it’s not even a very popular basket. Indeed, for very good scientific reasons, the overwhelming majority of the scientific community prefers gate model QCs to D-Wave’s QC. Gate model QCs are making fast, steady progress and the scientific theory predicts that they will be able to do things that a D-Wave machine can only dream of (like Shor’s algorithm, for example, or running my QC machine learning software 🙂 ).

All this talk about D-Wave gambling reminds me of the controversial prediction markets, which are praised in the book titled “The Wisdom of the Crowds“. Henning Dekant and Matthias Troyer already have a 2-part D-Wave bet going on. So let me start a little betting action here too. How about if I ask you to predict the future of D-Wave? (bookies throughout the world, take note)

July 9, 2014

In Love With Good Old Macsyma/Maxima

Filed under: Uncategorized — rrtucci @ 11:52 pm

I’d like to dedicate this blop post to extolling the many virtues of the Maxima symbolic manipulation program (SMP).

As I’ve mentioned in previous posts, I’m currently working on a paper that applies the representation theory of (mathematical) groups to quantum computing algorithms. I’m glad to say that this paper is almost finished! Most of the research part of it is done. Now I just have to write up my results. I’m planning to write a patent and an arXiv paper based on this work. I was able to do the research very quickly (for me). Did I cheat? No, but it feels like it. Normally, it would have taken me forever to do all the algebraic calculations associated with this project, but instead, I was able to do them very quickly because I was using a SMP, namely Maxima. I am immensely grateful to the hundreds of scientists that have contributed to Maxima.

The moral of this blog post is: Maxima is very well suited for doing certain aspects of quantum computing research, in particular stuff that involves symbolic manipulation of matrices. My current QC research involves a lot of this.

For my current paper, I had to be able to do all of the following symbolically (not numerically) without too much hassle and Maxima did not disappoint:

defining matrices, multiplying matrices, generating matrices that involve direct sums or Kronecker products of simpler matrices, finding the eigenvalues and eigenvectors of a matrix, finding the SVD (singular value decomposition) of a matrix, finding the CSD (Cosine-Sine Decomposition) of a matrix (the CSD is extremely useful when compiling a unitary matrix.), extracting submatices from a given matrix, etc., etc.

Maxima is easy to learn and the above tasks are easy to code. Furthermore, I have yet to find a serious bug in Maxima, or have it crash on me, and the beast is lightning fast at symbolics, which is what I needed for my current paper. On the other hand, Maxima is very limited in numerics. For that, use some other software. See numerics footnote below.

I was going to include in this blog post some examples of Maxima code, to whet your appetite, but I decided against it. What the heck, my next paper will include as an appendix a file with lots of Maxima code relevant to the paper. That should illustrate well enough to the budding quantum computerist how I use Maxima to help me write quantum computer code.

The Macsyma and Maxima codebases forked in 1982. Luckily, they haven’t diverged syntax-wise too much since then. After 1982, Macsyma added many new features but the syntax for all the basic stuff is still almost identical for both.

As for documentation available on the web, Maxima comes with a manual which is medium good (it’s not too bad but it could use more and better examples). I like very much the following tutorial originally produced by Symbolics and now available at the website of Richard J. Fateman (a Berkeley Univ. Prof. Emeritus, and one of the original authors of Macsyma while he was at MIT)

Users Guide, Second Edition

I found this tutorial to be really easy to read, fairly comprehensive, and with very good examples. This Macsyma “user’s guide” is not to be confused with the much longer but less friendly Macsyma “reference manual”. Whenever I copy a code line from the above mentioned Macsyma user’s guide and I find that the command line doesn’t work on Maxima, I then go to the Maxima manual and look in there for those changes, usually minor ones, that will make that particular code line more agreeable to Maxima.

Nowadays, most commercial SMPs have “notebooks” for storing and replaying your work. Maxima doesn’t have notebooks yet but I didn’t find this to be a big drawback for the work I was doing for my current paper. This is how I usually work with Maxima: I write all my commands first in a “project” txt file. Then I copy and paste those commands onto the Maxima command line. At the end of a Maxima session, I only save the project txt file. I begin each Maxima session by replaying all the commands in my latest project txt file.

Let me conclude this post with a brief overview of the bitter and turbulent history behind Macsyma/Maxima.

Brief History of Macsyma/Maxima

Macsyma, the first SMP, was developed at MIT from 1968 to 1982 with DOE and DARPA funding. After that, Macsyma passed into commercial hands (first Symbolics, then others). Between 1982-1999, many new features were added to the commercial version of Macsyma. Maxima is based on the last non-commercial, 1982 version of Macsyma. Maxima doesn’t include the features added to Macsyma from 1982-1999, but it has added a few features of its own, and it is available for a wide number of platforms as free, open, GPL-licensed software.

Although many have tried, nobody has ever been able to create a long-term stable and successful company based on Macsyma, this despite the fact that Macsyma software is quite robust and useful. This failure seems to have occurred for several reasons. One of them is that, as is well document throughout the Internet, there have been some bitter disagreements, infighting and nasty politics between many of Macsyma’s legal guardians (MIT academics, MIT management, Symbolics management, investors, DOE, etc.). Another reason seems to be that Stephen Wolfram ultimately proved himself to be a better businessman than those behind Macsyma. Macsyma can still be purchased for $500 today, but development of it has completely stopped. Nowadays, most of the commercial SMP market share belongs to Mathematica and Maple.

Wolfram didn’t like the fact that Macsyma was written in Lisp so he started writing his own SMP, originally called SMP, in the C language. His SMP was the basis of Mathematica. In the early days, Mathematica was weaker than Macsyma in symbolics but faster in numerics. Nowadays, Mathematica has a gazillion more bells and whistles than Macsyma/Maxima.

A footnote on Numerics

For numerics, you can use the free Lapack and non-free Matlab. The Octave program is a free partial clone of MatLab, certainly not as complete as MatLab, but good enough for doing basic stuff, and it has a syntax almost identical to Matlab’s for the basic suff. For my quantum computing software available at http://www.ar-tiste.com, I’ve used Lapack (in Qubiter, which is written in C++, but calls CLapack, the C translation of Lapack) and I’ve also used Octave (for my M-Fun For QC Progs)

July 7, 2014

1Qbit Company Switches Name to ZeroQbits

Filed under: Uncategorized — rrtucci @ 7:14 pm

The “1Qbit” company, founded about a year ago, is dedicated exclusively to writing software for D-Wave’s “quantum computer” (There is no indication from their flashy website that they are writing software for gate model QCs. Their main and only product so far is “Integer Optimization Toolbox”, which is useless for anything but a D-Wave computer). They’ve already opened 2 offices, one in Vancouver and another in Toronto, Canada. I’m sure an office in Silicon Valley will soon follow. Their publicity department has been quite the busy beaver. They’ve even produced their own page in Wikipedia at the tender age of one year.


All they need now is a machine that doesn’t decohere in a few nanoseconds and a software product that does something, anything, better than the classical competition. We’ll see what happens. Reminds me of mayflies (an ephemeral), an insect that gets to fly for approximately only one day, and then it dies.

July 3, 2014

Microsoft, the next Google?

Filed under: Uncategorized — rrtucci @ 6:47 pm

June must have been a rough month for the Google/D-Wave collaboration. A paper by Matthias Troyer et al which I quote below has been available at arXiv for a long time, in fact since January, but when it was published by Science magazine on June 19, all hell broke loose. The news media finally noticed the paper’s existence and came out in surprisingly large numbers to point out that the D-Wave computer had “flunked its first big test”. The Troyer et al paper concludes

“Using random spin glass instances as a benchmark, we find no evidence of quantum speedup when the entire data set is considered, and obtain inconclusive results when comparing subsets of instances on an instance-by-instance basis,”

Ouch, those words gotta hurt. Even cold hearted Scott Aaronson must feel a little sorry for D-Wave at this point.

To add Google-insult to Google-injury, the news media (including the NY Times, see here) has also started to trumpet the fact that, unlike doofus Google, Microsoft has been methodically subsidizing Gate Model QC research for about 15 years, and that research is beginning to bear fruit. Microsoft has a theoretical group called Station Q located at UCSB (Univ. of Calf. at Santa Barbara) doing research into anyonic QCs. They also have a small group of people writing generic gate model QC software. Also, for the last 15 years, they have been partially funding about a dozen experimental research groups throughout the world that are trying to build an anyonic QC. It appears that Robert Willet, working at the legendary Bell Labs in Murray Hill, N.J. (the little of it that still exists), is closest to building an anyonic qubit.

Last but not least, there is some evidence of civil Google-disobedience inside the very heart of Google country. Haker Dojo, the SETI Institute and the Googleplex (Google HQ) are all located in Mountain View CA (near San Francisco).

Haker Dojo has been hosting a MeetUp on quantum computing with roughly monthly events. For July 8, they have invited Nathan Wiebe from Microsoft, to speak about “Using quantum computers to learn physics”. This is a decidedly gate model talk, much to the consternation of Google. Like myself, Wiebe has written papers on using gate model QCs to do machine learning.

Meanwhile, SETI Institute is ringing another note of Google-disobedience by hosting on July 9 the first of what they hope will become an annual event, a FREE (as in beer) conference on Quantum Simulation. Their list of speakers for this year’s conference includes both gate-model and D-Wave partisans.

Blog at WordPress.com.

%d bloggers like this: