Jorge Cham, author of the PhD Comics strip just released an awesome animation about quantum computers, embedded below. On the accompanying post on the quantum frontiers blog you can read up a little bit about the story behind the clip. And if you are new to the PhD comics, take a look at the most popular PhD comics, to see why this little comic strip helps grad students world wide to see their daily routine a bit more lightheartedly.
The method I described last time started giving some very strange numbers for the hourly PSI yesterday June 21. Negative numbers! It turns out the 3-hour average PSI is not the average of the PSI for the last 3 hours. Huh?
Update: I see that someone else had the same idea here. Copying that plot, here is a plot of the 3 hour averages (in red) and 1 hour inferred averages (in blue) for June 20 from 6am to 5pm. Though the 3 hour average high of 371 at 1pm set a new record, the peak 1 hour average wasn’t that much worse than yesterday, about 467 by my calculation.
Singapore is obsessed with numbers these days. Well, one number in particular: the pollutant standards index or PSI. PSI measures the concentration of particles of diameter 10 micrometres or less in the air, that is about one-seventh the width of a human hair. A PSI value of 50 is fairly normal in a city these days. Over 100 starts to become noticeable and unhealthy, and above 200 feels like sitting next to a campfire with the wind blowing in your face.
The National Environment Agency hourly publishes PSI values for Singapore averaged over a 3 hour period. Because of fires in Sumatra the PSI levels have been elevated all this week, and last night (June 19) the PSI reading at 10pm hit a Singaporean record high of 321 (update: since broken with a 371 today). What I found even more remarkable, however, was that from 8pm to 9pm the 3-hour average jumped from 190 to 290. Since these readings share two data points, this means the jump from the 6pm reading to the 9pm reading must have been 300 points! The averages throughout the afternoon were mid-100′s, so it stands to reason that the 9pm reading was in the mid-400′s.
For some more fun, we can try to work out all the hourly averages from the 3 hour averages. Here is a table of the 3 hour averages from yesterday June 19 (currently found on the wikipedia page about PSI).
We have more unknowns than equations, so we will have to make some assumptions to calculate the hourly values. Notice that in the morning the PSI readings are fairly flat: 77, 78, 80 at 6am, 7am, and 8am. Here the 6am value is the average PSI in the time period 3-6am. The two assumptions I will make are that the hourly values from 3-4am and 4-5am were 77. Of course, this means that the 5-6am value is also 77 since we know that is the 3 hour average. With this assumption we can set up a set of linear equations and work out the hourly averages for the rest of the day.
For example, to figure out the hourly values from 6am to 11am we have the following matrix equation.
The hourly PSI average from 8-9pm, where we saw the spike in the 3-hour average, was a whopping 452. This is in line with the prediction from our simple reasoning above. The results are quite robust to assuming different values in the early morning—if the 3 hour averages are correct, it looks like in any case we must have gotten over 400 in the 8-9pm period!
This week Boxio et al. have put up test results on the arxiv on performing quantum annealing with more than one hundred qubits on the commercial D-Wave One device. Marketed as a quantum computer, the natural question to ask is “Will it blend?”. Or rather, is the device quantum or just a complex classical device? In their paper, the authors compare the experimental performance with a simulation of how a quantum device would perform. In addition they compare the behavior to optimized classical algorithms. The conclusion is interesting in two respects. First, the behavior of the current device seems to suggest it is a quantum device. And secondly, an improved version of the current architecture might be able to deliver real world speed ups in performing calculations with larger problem sizes. This would certainly be very interesting. Or in the words of the authors: “A quantum annealer showing better scaling than classical algorithms for these problem sizes would be an exciting breakthrough, validating the potential of quantum information processing to outperform its classical counterpart.”
Update June 5: There has been a lot of buzz about D-Wave in the aftermath of this preprint (and a related study). I highly recommend reading Scott Aaronson’s post about the topic, from which it seems that the jury is still out on whether the D-Wave architecture will provide real world benefits.
Update (22.05.2013): According to the Facebook page of the project, the game (now called “meQuanics”) will be released tomorrow.
Admittedly, one of the most common uses of computers is to play computer games, and in no small part did games influence the historic development of computer hardware. For example, the primary purpose of today’s high-end graphics cards is to compute the complex graphics effects of 3D games. Almost as an afterthought, it has been made possible to harness this brute computational power for productive purposes: Using frameworks such as OpenCL or CUDA, graphics cards can provide huge computational speedups in specific areas such as cryptography, molecular dynamics, fluid dynamics and distributed computing.
As far as quantum computers are concerned, some guys at the National Institute of Informatics (NII) in Tokyo believe that we can go in the opposite direction: By playing a computer game, we aid the development of error-correcting codes used for performing fault-tolerant quantum computations. The NII group is developing a mobile game whose goal is to compactify the surface codes used in topological quantum computation. To give an example, the quantum circuit on the left can be implemented by the surface code shown in the middle. The white and black “loops” in the surface code correspond to logical qubits, and quantum operations such as the CNOT-gate correspond to the braiding of the loops. For the finer details of how such codes correspond to the physical system, I refer to arXiv:1209.1441 and arXiv:1209.0510 (from which the three figures below were taken).
The code on the right shows a much compressed variant of the same underlying circuit, obtained from the code in the middle by a series of circuit-preserving transformations. One example of circuit-preserving transformations are transformations that preserve the topology of the loops and braids in the surface code, but there also exist non-trivial transformations, and the aim of the game is to minimize the volume of surface codes by performing such transformations. You can get an idea of how this will look like by viewing the trailer on the official homepage http://www.qubit-game.com or at YouTube.
Reducing the code size of important circuits, such as the ones performing purifications, has a huge potential to reduce the amount of resources (both in space and time) needed for the fault-tolerant implementation of quantum algorithms. However, little is known about the best strategies to compress surface codes, and this is where both fellow scientists and casual gamers step in: Their progress will be tracked online, thus enabling fancy stuff like score leaderboards and alerting the project managers when new compactification strategies have been discovered. Finding such strategies would be helpful for the development of compilers that automatically translate quantum algorithms into efficient surface codes with minimal overhead. Best of all, the discovery of particularly strong compactifications might earn you joint authorship in a scientific publication!
During a talk given by Simon Devitt, one of the project leaders, at CQT, I was able to try out a pre-release version of the game. The game is based on touch input, and will be available for Android- and Apple-powered phones and tablets. The 3D engine was fully working, and manipulations of the code were already possible. The roadmap is as follows: A closed beta version aimed towards the scientific community is scheduled to be released later this month. The feedback and “peer-review” from fellow researchers ensures that possible glitches (e.g. in the implementation of valid code transformations) can be spotted and fixed. Later this spring, the game will be released to the public, featuring a much refined user interface.
Using numerical simulations to explore the behavior of the quantum systems we study in the lab is a great tool for gaining more understanding. Especially if you want to go beyond the abstractions that most analytically solvable descriptions are. Unfortunately simulating quantum system on a classical computer scales very badly with the complexity of the system. Even simulating the single Barium ion trapped in the neighboring lab inside a high-finesse cavity with it’s full level structure and motional states may take too long on your personal computer for you to be patient enough to wait for the results. Not to mention to try and simulate the many neutral Rubidium atoms we trap in our lab.
The Quantum Toolbox in Python (QuTiP) is quite fast out of the box for these kinds of simulations, making use of core functions compiled to C code for speed and fully utilizing the multi-core processors present in all modern computers. But you still run into the problem that computations take too long, just for systems a little bit more complex. Being written in python, QuTiP makes it quite easy to offload some off the heavy computation onto a cluster in the cloud, for example via PiCloud. Recently I wrote a short guide on how-to get started with QuTiP on PiCloud, that shows you how easy it is to push the realm of doable computations quite bit further still. Since neither better software nor having a cloud based supercomputer at your fingertips gets around the fact that simulating quantum systems on a classical computer scales miserably, we have to make do with the imposed limits. Well that is until we have a quantum computer to run these simulations on…
This week I have been at the Dagstuhl castle in Germany for a workshop on “Communication complexity, linear optimization, and lower bounds for the nonnegative rank of matrices.” I always like coming to Dagstuhl, and this time it has been especially nice meeting many new people from the neighboring communities of combinatorial optimization and matrix theory.
Hidden on CQT’s website are some veritable gems which can be found among the colloquia given at CQT. All recent colloquia have been video taped, so that everyone can watch the talks online, and with the impressive list of speakers, you are bound to find plenty of talks of interest.
The colloquium is a monthly installment, where speakers have about an hour to give a more high level view of their field than during a normal talk. This makes for some really entertaining talks, and I can’t wait to watch the presentation that Eric Cornell gave last Wednesday again. In it, he managed to motivate his interest in making more precise measurements of the electron’s electric dipole moment by his desire to make supersymmetry the Paris Hilton of physics. That has gotten you interested? Unfortunately you have to wait a little bit more, because the video usually comes up about three weeks after the colloquium.
For now you can browse the colloquium archive. Last December for example, Avi Widgerson gave a highly enjoyable talk about randomness. His introduction to what randomness means from a computational scientist perspective is very interesting and the talk is easy to follow. Since it was given at CQT’s fifth birthday, the introduction of the speaker is a bit longer then usual and the talk starts for real only at the 4:30 mark of the video.
Update: Eric Cornell’s talk is now available online.
The Quantum Optics Toolbox in Python (qutip) is a marvellous tool for simulating open quantum systems. Firstly, Robert Johansson and Paul Nation have crafted a great tool which makes it easy to explore a wide range of quantum systems. And secondly, their tool builds upon the incredible scientific python stack.
Exploring quantum systems this way is a lot of fun, but knowing where to start is sometimes a bit daunting. Recently, Robert Johansson has put up some lecture notes on github in the IPython notebook format. These let you interactively explore the functionalities of qutip. Each lecture illustrates the behaviour of a particular quantum system. The great thing about the IPython notebook format is that these lecture notes are fully executable code, which means you can play around with the examples while working through them. An ideal way to dive right into simulating quantum systems with python!
If you are completely new to the whole python way of doing things, you find some tips on how to get started at the end of this post. To see what awaits you take a look at the static versions of his lecture notes below.
- Lecture 1 Jaynes Cumming model
- Lecture 2A Cavity Qubit Gates
- Lecture 2B Single Atom Lasing
- Lecture 3A Dicke model
- Lecture 3B Jaynes Cumming model with ultrastrong coupling
- Lecture 4 Correlation Functions
- Lecture 5 Parametric Amplifier
- Lecture 6 Quantum Monte Carlo Trajectories
On the github repository there are also PDF versions that you can download.
I was greatly saddened to hear of the death of Sean Barrett on Friday morning. It seems that a taxi he was traveling in was hit by a stolen SUV, while he was on his way to a conference in Perth. I was really shocked to hear the news.
Sean is pretty known within the quantum computing community for his work on entanglement generation and on measurement based computation. Indeed, one of his papers (on double-heralding entanglement generation) was a big influence on much of my PhD research.
Sean was a great guy, friendly, lively, and very smart. His death is a great loss to the community and my thoughts are with his family and loved ones at what must be a very difficult and painful time.