Serious users of Bayesian networks cannot avoid becoming experts in mathematical statistics. This year, with such an electrifying US presidential election, mathematical statistics has been frequently in the news. In “Two countries separated by a common language”, two Jedi Masters in their fields and bloggers par excellence, Terrence Tao and Andrew Gelman, explain polling statistics.
October 24, 2008
The Quantum Pontiff has commented with regard to the new zoo “Help squash the myth that all there is to quantum algorithms are the algorithms of Shor and Grover”
The bizarre thing about the algorithms in this zoo is that most of them (except for a rare few) solve problems that are not very general. Ouch!
Another bizarre feature about this zoo is that it ignores software completely, as if there were no connection between algorithms and software. (I’m not talking pseudo-code, I’m talking real code.) It seems some people in the quantum information field are wannabe string theorists. They think they are building a new mathematical theory, divorced from a new device, called a quantum computer, a device that will most certainly require software.
As a dumb user of computer algorithms, the best computer algorithm books I have ever read come with real code examples. My first introduction to computer algorithms was the truly wonderful and seminal book by Nijenhuis and Wilf, Combinatorial Algorithms for Computers and Calculators, which way back in the Stone Age of classical computers when it was first published, 1978, already provided full FORTRAN code for all its algorithms. Another book that illustrates my point is Numerical Recipes.
As a comparison to Jordan’s zoo, take a look at Steve Skiena’s excellent website, the Stony Brook Repository of computer algorithms, where the insoluble marriage of algorithms and software is well understood.
Please don’t tell me that quantum computing is too young to expect algorithms with software. I thought Pontiff’s and Mosca’s and Jordan’s point was that quantum computing has come a long way since Shor(1995) and Grover(1996). A long way without software? That’s funny.
October 17, 2008
I agree with these two lucid, sensible articles:
- Quantum Cryptography: As Awesome As It Is Pointless, by Bruce Schneier
- Quantum Snake Oil, by Matt Weir
(Bruce Schneier is a well-know cryptographer, computer security specialist, and writer. He is the author of several highly regarded books on computer security and cryptography. His undergraduate BS was in physics, so he is quite capable of understanding quantum cryptography in detail)
I too believe that quantum crypto, though beautiful physics, is pointless in practice. So why then do governments in the US and Europe continue to waste millions of dollars per year to fund this? One can only conclude that the scientists selling this to their governments are being disingenuous in their sales pitches. Too bad. The resources being taken up by quantum cryptography could be put to much better use within the meta field of quantum information. For instance, those resources could be put into trying harder to build a large scale quantum computer, and into writing software for it. It seems that some of the people doing quantum information research are not truly interested in building a quantum computer. Their main priority is to get research funding for something, no matter what, even if it’s not such a good topic and selling it requires plenty of fibbing. Often, rather than changing their research topics over time, they want to continue to get research funding forever, for the same old topic which has lost its luster long ago.
October 13, 2008
I am currently busily occupied writing a paper on using a quantum computer to sample classical Bayesian networks. The new paper will extend my previous results reported here.
Wikipedia has an excellent collection of articles on the subject I am working on, under the category “Monte Carlo methods”.
I think when tackling any subject for the first time, it’s important to make a mental map of the “lay of the land”. Here is a topics tree of the subject of Monte Carlo sampling as I see it:
- univariate (single random variable, single node Bayesian net)
- inverse transform method
- ARM (acceptance rejection method)
- multi-variate (multiple random variables, multi-node Bayesian net)
- Importance Sampling
- Rejection Sampling
- Likelihood weighted sampling
- MCMC (Markov Chain Monte Carlo)
- Simulated Annealing
- Importance Sampling
October 4, 2008
Consider the following quantum state (expressed in Dirac notation):
where 0,1,2 label qubits. Suppose you wanted to portray this state graphically, as a “quantum circuit diagram”.
In my papers, I always draw quantum circuit diagrams with time pointing from right to left. I call this the Dirac convention. Thus, I would portray the state of Eq.(1) as
Most scientists in the field of quantum computing draw their diagrams with time pointing from left to right, in proper Christian order. I call this the Quayle convention (in honor of Dan Quayle, U.S. Vice President from 1989 to 1993). Thus, Dan Quayle would portray the state of Eq.(1) as
Be forewarned that users of the Quayle convention consider the Dirac convention uncouth and un-Christian (used by sinful Jews and Muslims in their uncivilized Hebrew and Arabic languages).
It’s also interesting to ponder how Reverend Bayes would have portrayed this state. He would, no doubt, have done so as follows, in what I like to call the Bayes convention (i.e., as a quantum Bayesian network):
Dan Quayle Holding Grenade Launcher
Date Photographed: June 13, 1989
Location: El Salvador
Photographer: Bill Gentile
(higher resolution available at Corbis)